Data Analysis and Seismogram Interpretation

11
Data Analysis and Seismogram

Interpretation
Peter Bormann, Klaus Klinge and Siegfried Wendt
11.1 Introduction
This Chapter deals with seismogram analysis and extraction of seismic parameter values for
data exchange with national and international data centers, for use in research and last, but not
least, with writing bulletins and informing the public about seismic events. It is written for
training purposes and for use as a reference source for seismologists at observatories. It
describes the basic requirements in analog and digital routine observatory practice i.e., to:
• recognize the occurrence of an earthquake in a record;

• identify and annotate the seismic phases;
• determine onset time and polarity correctly;
• measure the maximum ground amplitude and related period;
• calculate slowness and azimuth;
• determine source parameters such as the hypocenter, origin time, magnitude, source
mechanism, etc..
In modern digital observatory practice these procedures are implemented in computer

programs. Experience, a basic knowledge of elastic wave propagation (see Chapter 2), and the
available software can guide a seismologist to analyze large amounts of data and interpret
seismograms correctly. The aim of this Chapter is to introduce the basic knowledge, data,
procedures and tools required for proper seismogram analysis and phase interpretation and to
present selected seismogram examples.
Seismograms are the basic information about earthquakes, chemical and nuclear explosions,
mining-induced earthquakes, rock bursts and other events generating seismic waves.
Seismograms reflect the combined influence of the seismic source (see Chapter 3), the
propagation path (see Chapter 2), the frequency response of the recording instrument (see 4.2
and 5.2), and the ambient noise at the recording site (e.g., Fig. 7.32). Fig. 11.1 summarizes
these effects and their scientific usefulness. Accordingly, our knowledge of seismicity, Earth's
structure, and the various types of seismic sources is mainly the result of analysis and
interpretation of seismograms. The more completely we quantify and interpret the
seismograms, the more fully we understand the Earth's structure, seismic sources and the
underlying causing processes.
Fig. 11.1 Different factors/sub-systems (without seismic noise) which influence a seismic record (yellow boxes)
and the information that can be derived from record analysis (blue boxes).
Seismological data analysis for single stations is nowadays increasingly replaced by network
(see Chapter 8) and array analysis (see Chapter 9). Array-processing techniques have been
developed for more than 20 years. Networks and arrays, in contrast to single stations, enable
better signal detection and source location. Also, arrays can be used to estimate slowness and
azimuth, which allow better phase identification. Further, more accurate magnitude values can
be expected by averaging single station magnitudes and for distant sources the signal
coherency can be used to determine onset times more reliably. Tab. 11.1 summarizes basic
characteristics of single stations, station networks and arrays. In principle, an array can be
used as a network and in special cases a network can be used as an array. The most important
differences between networks and arrays are in the degree of signal coherence and the data
analysis techniques used.
Like single stations, band-limited seismometer systems are now out-of-date and have a
limited distribution and local importance only. Band-limited systems filter the ground
motion. They distort the signal and may shift the onset time and reverse polarity (see 4.2).
Most seismological observatories, and especially regional networks, are now equipped with
broadband seismometers that are able to record signal frequencies between about 0.001 Hz
and 50 Hz. The frequency and dynamic range covered by broadband recordings are shown in
Fig. 11.2 and in Fig. 7.48 of Chapter 7 in comparison with classical band-limited analog
recordings of the Worldwide Standard Seismograph Network (WWSSN).
Tab. 11.1 Short characteristic of single stations, station networks and arrays.
11.1 Introduction
Single station Classical type of seismic station with its own data processing. Event
location only possible by means of three-component records.
Station network Local, regional or global distribution of stations that are as identical as
possible with a common data center (see Chapter 8). Event location is
one of the main tasks.
Seismic array Cluster of seismic stations with a common time reference and uniform
instrumentation. The stations are located close enough to each other in
space for the signal waveforms to be correlated between adjacent
sensors (see Chapter 9). Benefits are:
• extraction of coherent signals from random noise;
• determination of directional information of approaching
wavefronts (determination of backazimuth of the source);
• determination of local slowness and thus of epicentral
distance of the source.
Fig. 11.2 Frequency range of seismological interest.
A number of these classical seismograph systems are still in operation at autonomous single
stations in many developing countries and in the former Soviet Union. Also, archives are
filled with analog recordings of these systems, which were collected over many decades.
These data constitute a wealth of information most of which has yet to be fully analyzed and
scientifically exploited. Although digital data are superior in many respects, both for
advanced routine analysis and even more for scientific research, it will be many years or even
decades of digital data acquisition before one may consider the bulk of these old data as no
longer needed. However, for the rare big and thus unique earthquakes, and for earthquakes in
areas with low seismicity rates but significant seismic risk, the preservation and
comprehensive analysis of these classical and historic seismograms will remain of the utmost
importance for many years.
More and more old analog data will be reanalyzed only after being digitized and by using
similar procedures and analysis programs as for recent original digital data. Nevertheless,
station operators and analysts should still be in a position to handle, understand and properly
analyze analog seismograms or plotted digital recordings without computer support and with
only modest auxiliary means. Digital seismograms are analyzed in much the same way as
classical seismograms (although with better and more flexible time and amplitude resolution)
except that the digital analysis uses interactive software which makes the analysis quicker and
easier, and their correct interpretation requires the same knowledge of the appearance of
seismic records and individual seismic phases as for analog data. The analyst needs to know
the typical features in seismic records as a function of distance, depth and source process of
the seismic event, their dependence on the polarization of the different types of seismic waves
and thus of the azimuth of the source and the orientation of sensor components with respect to
it. He/she also needs to be aware of the influence of the seismograph response on the
appearance of the record. Without this solid background knowledge, phase identifications and
parameter readings may be rather incomplete, systematically biased or even wrong, no matter
what kind of sophisticated computer programs for seismogram analysis are used.
Therefore, in this Chapter we will deal first with an introduction to the fundamentals of
seismogram analysis at single stations and station networks, based on analog data and
procedures. Even if there is now less and less operational need for this kind of instruction and
training, from an educational point of view its importance can not be overemphasized. An
analyst trained in comprehensive and competent analysis of traditional analog seismic
recordings, when given access to advanced tools of computer-assisted analysis, will by far
outperform any computer specialist without the required seismological background
knowledge.
Automated phase identification and parameter determination is still inferior to the results
achievable by well-trained man-power. Therefore, automated procedures are not discussed in
this Manual although they are being used more and more at advanced seismological
observatories as well as at station networks (see Chapter 8) and array centers (see Chapter 9).
The Manual chiefly aims at providing competent guidance and advice to station operators and
seismologists with limited experience and to those working in countries which lack many
specialists in the fields which have to be covered by observatory personnel. On the other
hand, specialists in program development and automation algorithms sometimes lack the
required seismological knowledge or the practical experience to produce effective software
for observatory applications. Such knowledge and experience, however, is an indispensable
requirement for further improvement of computer procedures for automatic data analysis,
parameter determination and source location in tune also with older data and established
standards. In this sense, the Manual also addresses the needs of this advanced user
community.
11.1 Introduction
Accordingly, we first give a general introduction to routine seismogram interpretation of

analog recordings at single stations and small seismic networks. Then we discuss both the
similarities and the principal differences when processing digital data. The basic requirements
for parameter extraction, bulletin production as well as parameter and waveform data
exchange are also outlined. In the sub-Chapter on digital seismogram analysis we discuss in
more detail problems of signal coherence, the related different procedures of data processing
and analysis as well as available software for it. The majority of record examples from
Germany has been processed with the program Seismic Handler (SHM) developed by K.
Stammler which is used for seismic waveform retrieval and data analysis. This program and
descriptions are available via http://www.szgrf.bgr.de/sh-doc/index.html. Reference is made,
however, to other analysis software that is widely used internationally (see 11.4).
Typical examples of seismic records from different single stations, networks and arrays in
different distance ranges (local, regional and teleseismic) and at different source depth are
presented, mostly broadband data or filtered records derived therefrom. A special section is
dedicated to the interpretation of seismic core phases (see 11.5.2.4 and 11.5.3). Since all
Chapter authors come from Germany, the majority of records shown has unavoidably been
collected at stations of the German Regional Seismic Network (GRSN) and of the Gräfenberg
array (GRF). Since all these stations record originally only velocity-broadband (BB-velocity)
data, all examples shown from GRSN/GRF stations of short-period (SP), long-period (LP) or
BB-displacement seismograms corresponding to Wood-Anderson, WWSSN-SP, WWSSN-
LP, SRO-LP or Kirnos SKD response characteristics, are simulated records. Since their
appearance is identical with respective recordings of these classical analog seismographs this
fact is not repeatedly stated throughout this Chapter and its annexes. The location and
distribution of the GRSN and GRF stations is depicted in Fig. 11.3a. while Fig. 11.3b shows
the location of the events for which records from these stations are presented. Users of this
Chapter may feel that the seismograms presented by the authors are too biased towards
Europe. Indeed, we may have overlooked some important aspects or typical seismic phases
which are well observed in other parts of the world. Therefore, we invite anybody who can
present valuable complementary data and explanations to submit them to the Editor of the
Manual so that they can be integrated into future editions of the Manual.
For routine analysis and international data exchange a standard nomenclature of seismic
phases is required. The newly elaborated draft of a IASPEI Standard Seismic Phase List is
given in IS 2.1, together with ray diagrams for most phases. This new nomenclature partially
modifies and completes the earlier one published in the last edition of the Manual of
Seismological Observatory Practice (Willmore, 1979) and each issue of the seismic Bulletins
of the International Seismological Centre (ISC). It is more in tune than the earlier versions
with the phase definitions of modern Earth and travel-time models (see 2.7) and takes full
advantage of the newly adopted, more flexible and versatile IASPEI Seismic Format (ISF; see
10.2.5) for data transmission, handling and archiving.
The scientific fundamentals of some of the essential subroutines in any analysis software are
separately treated in Volume 2, Annexes (e.g., IS 11.1 or PD 11.1). More related Information
Sheets and Program Descriptions may be added in the course of further development of this
Manual.
a)
b)
Fig. 11.3 a) Stations of the German Regional Seismological Network (GRSN, black
triangles) and the Graefenberg-Array (GRF, green dots); b) global distribution of epicenters
of seismic events (red dots: underground nuclear explosions; yellow dots: earthquakes) for
which records from the above stations will be presented in Chapter 11 and DS 11.1-11.4.
11.2 Criteria and parameters for routine seismogram analysis

11.2.1 Record duration and dispersion
The first thing one has to look for when assessing a seismic record is the duration of the
signal. Due to the different nature and propagation velocity of seismic waves and the different
propagation paths taken by them to a station, travel-time differences between the main wave
groups usually grow with distance. Accordingly, the record spreads out in time. The various
body-wave groups show no dispersion, so their individual duration remains more or less
constant, only the time-difference between them changes with distance (see Fig. 2.48). The
time difference between the main body-wave onsets is roughly < 3 minutes for events at
distances D < 10°, < 16 min for D < 60°, < 30 min for D < 100° and < 45 min for D < 180°
(see Fig. 1.2).
In contrast to body waves, velocity of surface waves is frequency dependent and thus surface
waves are dispersed. Accordingly, depending on the crustal/mantle structure along the
propagation path, the duration of Love- and Rayleigh-wave trains increases with distance. At
D > 100° surface wave seismograms may last for an hour or more (see Fig. 1.2), and for really
strong events, when surface waves may circle the Earth several times, their oscillations on
sensitive long-period (LP) or broadband (BB) records may be recognizable over 6 to 12 hours
(see Fig. 2.19). Even for reasonably strong regional earthquakes, e.g., Ms ≈ 6 and D ≈ 10°, the
oscillations may last for about an hour although the time difference between the P and S onset
is only about 2 min and between P and the maximum amplitude in the surface wave group
only 5-6 min.
Finally, besides proper dispersion, scattering may also spread wave energy. This is
particularly true for the more high-frequency waves traveling in the usually heterogeneous
crust. This gives rise to signal-generated noise and coda waves. Coda waves follow the main
generating phases with exponentially decaying amplitudes. The coda duration depends
mainly on the event magnitude (see Figure 1b in DS 11.1) and only weakly on epicentral
distance (see Figure 2 in EX 11.1). Thus, duration can be used for calculating magnitudes Md
(see 3.2.4.3).
In summary, signal duration, the time difference between the Rayleigh-wave maximum and
the first body-wave arrival (see Table 5 in DS 3.1) and in particular the time span between the
first and the last recognized body-wave onsets before the arrival of surface waves allow a first
rough estimate, whether the earthquake is a local, regional or teleseismic one. This rough
classification is a great help in choosing the proper approach, criteria and tools for further
more detailed seismogram analysis, source location, and magnitude determination.
11.2.2 Key parameters: Onset time, amplitude, period and polarity

Onset times of seismic wave groups, first and foremost of the P-wave first arrival, when
determined at many seismic stations at different azimuth and at different distance, are the key
input parameter for the location of seismic events (see IS 11.1). Travel times published in
travel-time tables (such as Jeffreys and Bullen, 1940; Kennett, 1991) and travel-time curves,
such as those shown in Figs. 2.40 and 2.50 or in the overlays to Figs. 2.47 and 2.48, have been
derived either from observations or Earth models. They give, as a function of epicentral
distance D and hypocentral depth h, the differences between onset times tox of the respective
seismic phases x and the origin time OT of the seismic source. Onset times mark the first
energy arrival of a seismic wave group. The process of recognizing and marking a wave onset
and of measuring its onset time is termed onset time picking. The recognition of a wave onset
largely depends on the spectral signal-to-noise-ratio (SNR) for the given waveform as a whole
and the steepness and amplitude of its leading edge. Both are controlled by the shape and
bandwidth of the recording seismograph or filter (see Figs. 4.9 to 4.13). It is a classical
convention in seismological practice to classify onsets, as a qualitative measure for the
reliability of their time-picking, as either impulsive (i) or emergent (e). These lower case
letters i or e are put in front of the phase symbol. Generally, it is easier to recognize and
precisely pick the very first arrival (usually a P wave) on a seismogram than later phases that
arrive within the signal-generated noise coda of earlier waves.
The relative precision with which an onset can be picked largely depends on the factors
discussed above, but the absolute accuracy of onset-time measurement is controlled by the
available time reference. Seismic body-wave phases travel rather fast. Their apparent
velocities at the surface typically range between about 3 km/s and nearly 100 km/s (at the
antipode the apparent velocity is effectively infinite). Therefore, an absolute accuracy of
onset-time picking of less than a second and ideally less than 0.1 s is needed for estimating
reliable epicenters (see IS 11.1) and determining good Earth models from travel-time data.
This was difficult to achieve in earlier decades when only mechanical pendulum clocks or
marine chronometers were available at most stations. They have unavoidable drifts and could
rarely be checked by comparison with radio time signals more frequently than twice a day.
Also, the time resolution of classical paper or film records is usually between 0.25 to 2 mm
per second, thus hardly permitting an accuracy of time-picking better than a second. In
combination with the limited timing accuracy, the reading errors at many stations of the
classical world-wide network, depending also on distance and region, were often two to three
seconds (Hwang and Clayton, 1991). However, this improved since the late 1970s with the
availability of very-low frequency and widely received time signals, e.g., from the DCF and
Omega time services, and recorders driven with exactly 50 Hz stabilized alternating current.
Yet, onset-time reading by human eye from analog records with minute marks led to
sometimes even larger errors, a common one being the ± 1 min for the P-wave first arrival.
This is clearly seen in Fig. 2.46 (left), which shows the travel-time picks collected by the ISC
from the world-wide seismic station reports between 1964 and 1987. Nowadays, atomic clock
time from the satellite-borne Global Positioning System (GPS) is readily available in nearly
every corner of the globe. Low-cost GPS receivers are easy to install at both permanent and
temporary seismic stations and generally affordable. Therefore the problem of unreliable
absolute timing should no longer exist. Nevertheless, also with high resolution digital data
and exact timing now being available it is difficult to decide on the real signal onset, even for
sharp P from explosions. Douglas et al. (1997) showed that the reading errors have at best a
standard deviation between 0.1 and 0.2 s. However, human reading errors no longer play a
role when digital data are evaluated by means of seismogram analysis software which
automatically records the time at the positions where onsets have been marked with a cursor.
Moreover, the recognizability of onsets and the precision of time picks can be modified easily
within the limits which are set by the sampling rate and the dynamic range of recording. Both
the time and amplitude scales of a record can be compressed or expanded as needed, and task-
dependent optimal filters for best phase recognition can be easily applied.
Fig. 11.4 shows such a digital record with the time scale expanded to 12 mm/s. The onset time
can be reliably picked with an accuracy of a few tenths of a second. This P-wave first arrival
has been classified as an impulsive (i) onset, although it looks emergent in this particular plot.
But by expanding the amplitude scale also, the leading edge of the wave arrival becomes
steeper and so the onset appears impulsive. This ease with which digital records can be
manipulated largely eliminates the value of qualitative characterization of onset sharpness by
either i or e. Therefore, in the framework of the planned but not yet realized International
Seismological Observing Period (ISOP), it is proposed instead to quantify the onset-time
reliability. This could be done by reporting, besides the most probable or interpreter-preferred
onset time, the estimated range of uncertainty by picking the earliest (tox- ) and latest possible
onset time (tox+) for each reported phase x, and of the first arrival in particular (see Fig. 11.6 ).
Fig. 11.4 First motion onset times, phase and polarity readings (c – compression; d –
dilatation), maximum amplitude A and period T measurements for a sharp (i - impulsive)
onset of a P wave from a Severnaya Zemlya event of April 19, 1997, recorded by a broadband
three-component single station of the Gräfenberg Array, Germany.
Whereas the quality, quantity and spatial distribution of reported time picks largely controls
the precision of source locations (see IS 11.1), the quality and quantity of amplitude readings
for identified specific seismic phases determine the representativeness of classical event
magnitudes. The latter are usually based on readings of maximum ground-displacement and
related periods for body- and surface-wave groups (see 3.2). For symmetric oscillations
amplitudes should be given as half peak-to-trough (double) amplitudes. The related periods
should be measured as the time between neighboring peaks (or troughs) of the amplitude
maximum or by doubling the time difference between the maximum peak and trough (see Fig.
11.4 and Fig. 3.9). Only for highly asymmetric wavelets should the measurement be made
from the center line to the maximum peak or trough (see Fig. 3.9b). Some computer programs
mark the record cycle from which the maximum amplitude A and the related T have been
measured (see Figures 3 and 4 in EX 3.1).
Note that the measured maximum trace amplitudes in a seismic record have to be corrected
for the frequency-dependent magnification of the seismograph to find the “true” ground-
motion amplitude, usually given in nanometers (1 nm = 10-9m) or micrometers (1 µm = 10-
6
m), at the given period. Fig. 3.11 shows a few typical displacement amplification curves of
standard seismographs used with paper or film records. For digital seismographs, instead of
displacement magnification, the frequency dependent resolution is usually given in units of
nm/counts, or in nm s-1/count for ground velocity measurements. Note that both record
amplitudes and related dominating periods do not only depend on the spectrum of the arriving
waves but are mainly controlled by the shape, center frequency and bandwidth of the
seismograph or record filter response (see Fig. 4.13). Also, the magnifications given in the
seismograph response curves are strictly valid only for steady-state harmonic oscillations
without any transient response. The latter, however, might be significant when narrow-band
seismographs record short wavelets of body waves. Signal shape, amplitudes and signal
duration are then heavily distorted (see Figs. 4.10 and 4.17). Therefore, we have written
“true” ground motion in quotation marks. Scherbaum (2001) gives a detailed discussion of
signal distortion which is not taken into account in standard magnitude determinations from
band-limited records. However, signal distortion must be corrected for in more advanced
digital signal analysis for source parameter estimation. The distortions are largest for the very
first oscillation(s) and they are stronger and longer lasting the narrower the recording
bandwidth (see 4.2.1 and 4.2.2). The transient response decays with time, depending also on
the damping of the seismometer. It is usually negligible for amplitude measurements on
dispersed teleseismic surface wave trains.
To calculate ground motion amplitudes from record amplitudes, the frequency-dependent

seismometer response and magnification have to be known from careful calibration (see 5.8).
Analog seismograms should be clearly annotated and relate each record to a seismometer with
known displacement magnification. For digital data, the instrument response is usually
included in the header information of each seismogram file or given in a separate file that is
automatically linked when analyzing data files. As soon as amplitudes and associated periods
are picked in digital records, most software tools for seismogram analysis calculate
instantaneously the ground displacement or ground velocity amplitudes and write them in
related parameter files.
Another parameter which has to be determined (if the signal-to-noise-ratio permits) and
reported routinely is the polarity of the P-wave first motion in vertical component records.
Reliable observations of the first motion polarity at stations surrounding the seismic source in
different directions allows the derivation of seismic fault-plane solutions (see 3.4 and EX 3.2).
The wiring of seismometer components has to be carefully checked to assure that
compressional first arrivals (c) appear on vertical-component records as an upward motion (+)
while dilatational first arrivals (d) are recorded as a downward first half-cycle (-). The
conventions for horizontal component recordings are + (up) for first motions towards N and
E, and – (down) for motions towards S and W. These need to be taken into account when
determining the backazimuth of the seismic source from amplitude and polarity readings on
3-component records (see EX 11.2, Figure 1). However, horizontal component polarities are
not considered in polarity-based fault-plane solutions and therefore not routinely reported to
data centers. Fig. 11.4 shows a compressional first arrival.
One should be aware, however, that narrow-band signal filtering may reduce the first-motion
amplitude by such a degree that its polarity may no longer be reliably recognized or may even
become lost completely in the noise (see Figs. 4.10 and 4.13). This may result in the wrong
polarity being reported and hence erroneous fault-plane solutions. Since short-period (SP)
records usually have a narrower bandwidth than medium- to long-period or even broadband
records, one should differentiate between first-motion polarity readings from SP and LP/BB
records. Also, long-period waves integrate over much of the detailed rupture process and so
should show more clearly the overall direction of motion which may not be the same as the
first-motion arrival in SP records which may be very small. Therefore, when reporting
polarities to international data centers one should, according to recommendations in 1985 of
the WG on Telegrafic Formats of the IASPEI Commission on Practice, unambiguously
differentiate between such readings on SP (c and d) and those on LP and BB records,
respectively (u for “up” = compression and r for “rarefaction” = dilatation). Note, however,
that reliable polarity readings are only possible on BB records!
Fig. 11.5 WWSSN-SP vertical-component records of GRSN stations for the same event as in
Fig. 11.4. While the P-wave amplitudes vary significantly within the network, the first-motion
polarity remains the same.
11.2.3 Advanced wavelet parameter reporting from digital records

The parameters discussed in 11.2.1 have been routinely reported over the decades of analog
recording. Digital records, however, allow versatile signal processing so that additional
wavelet parameters can be measured routinely. Such parameters may provide a much deeper
insight into the seismic source processes and the seismic moment release. Not only can onset
times be picked but their range of uncertainty can also be marked. Further, for a given wave
group, several amplitudes and related times may be quickly measured and these allow
inferences to be drawn on how the rupture process may have developed in space and time.
Moreover, the duration of a true ground displacement pulse tw and the rise time tr to its
maximum amplitude contain information about size of the source, the stress drop and the
attenuation of the pulse while propagating through the Earth. Integrating over the area
underneath a displacement pulse allows to determine its signal moment ms which is,
depending on the bandwidth and corner period of the recording, related to the seismic
moment M0 (Seidl and Hellweg, 1988). Finally, inferences on the attenuation and scattering
properties along the wave path can be drawn from the analysis of wavelet envelopes.
Fig. 11.6 depicts various parameters in relation to different seismic waveforms. One has to be
aware, however, that each of these parameters can be severely affected by the properties of
the seismic recording system (see Fig. 4.17 and Scherbaum, 1995 and 2001). Additionally,
one may analyze the signal-to-noise ratio (SNR) and report it as a quantitative parameter for
characterizing signal strength and thus of the reliability of phase and parameter readings. This
is routinely done when producing the Reviewed Event bulletin (RED) of the International
Data Centre (IDC) in the framework of the CTBTO. The SNR may be either given as the ratio
between the maximum amplitude of a considered seismic phase to that of the preceding
ambient or signal-generated noise, or more comprehensively by determining the spectral
SNR (see Fig. 11.47).
Fig. 11.6 Complementary signal parameters such as multiple wavelet amplitudes and related
times, rise-time tr of the displacement pulse, signal moment ms and wavelet envelope (with
modification from Scherbaum, Of Poles and Zeros, Fig. 1.9, p. 10,  2001; with permission
of Kluwer Academic Publishers).
Although these complementary signal parameters could be determined rather easily and
quickly by using appropriate software for signal processing and seismogram analysis, their
measurement and reporting to data centers is not yet common practice. It is expected,
however, that the recently introduced more flexible formats for parameter reporting and
storage (see ISF, 10.2.5), in conjunction with e-mail and internet data transfer, will pave the
way for their routine reporting.
11.2.4 Criteria to be used for phase identification

11.2.4.1 Travel time and slowness
As outlined in Chapter 2, travel times of identified seismic waves are not only the key
information for event location but also for the identification of seismic wave arrivals and the
determination of the structure of the Earth along the paths which these waves have traveled.
The same applies to the horizontal component sx of the slowness vector s. The following
relations hold:
sx = dt/dD = p = 1/vapp
were vapp is the apparent horizontal velocity of wave propagation, dt/dD the gradient of the
travel-time curve t(D) in the point of observation at distance D, and p is the ray parameter.
Due to the given structure of the Earth, the travel-time differences between various types of
seismic waves vary with distance in a systematic way. Therefore, differential travel-time
curves with respect to the P-wave first arrival (see Figure 4 in EX 11.2) or absolute travel-
time curves with respect to the origin time OT (see Figure 4 in EX 11.1 or overlay to Fig.
2.48) are the best tools to identify seismic waves on single station records. This is done by
matching as many of the recognizable wave onsets in the record as possible with travel-time
curves for various theoretically expected phases at epicentral distance D.
Make sure that the plotted t(D)-curves have the same time-resolution as your record and
investigate the match at different distances. Relative travel-time curves thus allow not only
the identification of best matching phases but also the distance of the station from the
epicenter of the source to be estimated. Note, however, that from certain distance ranges the
travel-time curves of different types of seismic waves (see Figure 4 in EX 11.2) are close to
each other, or even overlap, for example for PP and PcP between about 40° and 50° (see
Figure 6a in DS 11.2) and for S, SKS and ScS between 75° and 90° (see Fig. 11.7 and 11.54).
Proper phase identification then requires additional criteria besides travel-time differences to
be taken into account (see 11.2.4.2 to 11.2.4.4). Select the most probable distance by taking
these additional criteria into account. Absolute travel-time curves allow also the origin time to
be estimated (see exercises EX 11.1 and EX 11.2).
Fig. 11.7 Example of long-period horizontal component seismogram sections from a deep-
focus earthquake in the Sea of Okhotsk (20.04.1984, mb = 5.9, h = 588 km), recorded at the
stations RSSD, RSNY, and RSCP, respectively, in the critical distance range of overlapping
travel-time branches of S, SKS and ScS. Because of the large focal depth the depth phase sS
is clearly separated in time.
Note, however, that the travel-time curves shown in the overlays to Figs. 2.47 and 2.48, or
those given in EX 11.1 and EX 11.2, are valid for near-surface sources only. Both absolute
and (to a lesser extent) relative travel times change with source depth (see IASPEI 1991
Seismological Tables, Kennett, 1991) and, in addition, depth phases may appear (see Fig.
2.43 and Table 1 in EX 11.2). Note also that teleseismic travel-time curves (D > (17)20°) vary
little from region to region. Typically, the theoretical travel times of the main seismic phases
deviate by less than 2 s from those observed (see Fig. 2.52). In contrast, local/regional travel-
time curves for crustal and uppermost mantle phases may vary strongly from region to region.
This is due to the pronounced lateral variations of crustal thickness and structure (see Fig.
2.10), age, and seismic wave velocities in continental and oceanic areas. This means
local/regional travel-time curves have to be derived for each region in order to improve phase
identification and estimates of source distance and depth.
Often, rapid epicentre and/or source depth estimates are already available from data centers
prior to detailed record analysis at a given station. Then modern seismogram analysis
software such as SEISAN (Havskov, 1996; Havskov and Ottemöller, 1999), SEIS89
(Baumbach, 1999), GIANT (Rietbrock and Scherbaum, 1998) or Seismic Handler (SH and
SHM) (Stammler, http://www.szgrf.bgr.de/sh-doc/index.html) allow the theoretically
expected travel times for all main seismic phases to be marked on the record. This eases phase
identification. An example is shown in Fig. 11.13 for a record analyzed with Seismic Handler.
However, theoretically calculated onset-times based on a global average model should

only guide the phase identification but not the picking of onsets! Be aware that one of the
major challenges for modern global seismology is 3-D tomography of the Earth. What are
required are the location and the size of anomalies in wave velocity with respect to the global
1-D reference model. Only then will material flows in the mantle and core (which drive plate
tectonics, the generation of the Earth's magnetic field and other processes) be better
understood. Station analysts should never trust the computer generated theoretical onset times
more than the ones that they can recognize in the record itself. For Hilbert transformed phases
(see 2.5.4.3) onset times are best read after filtering to correct for the transforming. Without
unbiased analyst readings we will never be able to derive improved models of the
inhomogeneous Earth. Moreover, the first rapid epicenters, depths and origin-times published
by the data centers are only preliminary estimates and are usually based on first arrivals only.
Their improvement, especially with respect to source depth, requires more reliable onset-time
picks, and the identification of secondary (later) arrivals (see Figure 7 in IS 11.1).
At a local array or regional seismic network center both the task of phase identification and of
source location is easier than at a single station because local or regional slowness can be
measured from the time differences of the respective wave arrivals at the various stations (see
9.4, 9.5, 11.3.4 and 11.3.5). But even then, determining D from travel-time differences
between P or PKP and later arrivals can significantly improve the location accuracy. This is
best done by using three-component broadband recordings from at least one station in the
array or network. The reason this is recommended is that travel-time differences between first
and later arrivals vary much more rapidly with distance than the slowness of first arrivals. On
the other hand, arrays and regional networks usually give better control of the backazimuth of
the source than 3-component recordings (see 11.2.4.3), especially for low-magnitude events.
11.2.4.2 Amplitudes, dominating periods and waveforms
Amplitudes of seismic waves vary with distance due to geometric spreading, focusing and
defocusing caused by variations in wave speed and attenuation. To correctly identify body-
wave phases one has first to be able to differentiate between body- and surface-wave groups
and then estimate at least roughly, whether the source is at shallow, intermediate or rather
large depth. At long range, surface waves are only seen on LP and BB seismograms. Because
of their 2D propagation, geometrical spreading for surface waves is less than for body waves
that propagate 3-D. Also, because of their usually longer wavelength, surface waves are less
attenuated and affected less by small-scale structural inhomogeneities than body waves.
Therefore, on records of shallow seismic events, surface-wave amplitudes dominate over
body-wave amplitudes (see Figs. 11.8 and 11.9) and show less variability with distance (see
Fig. 3.13). This is also obvious when comparing the magnitude calibration functions for body
and surface waves (see figures and tables in DS 3.1).
Fig. 11.8 Three-component BB-velocity record at station MOX of a mine collapse in

Germany; (13 March 1989; Ml = 5.5) at a distance of 112 km and with a backazimuth of
273°. Note the Rayleigh surface-wave arrival LR with subsequent normal dispersion.
Fig. 11.9 T-R-Z rotated three-component seismogram (SRO-LP filter) from an earthquake
east of Severnaya Zemlya (19 April 1997, D = 46.4°, mb = 5.8, Ms = 5.0). The record shows
P, S, SS and strong Rayleigh surface waves with clear normal dispersion. The surface wave
maximum has periods of about 20 s. It is called an Airy-phase and corresponds to a minimum
in the dispersion curve for continental Rayleigh waves (see Fig. 2.9).
However, as source depth increases, surface-wave amplitudes decrease relative to those of
body waves, the decrease being strongest for shorter wavelengths. Thus, the surface waves
from earthquakes at intermediate (> 70 km) or great depth (> 300 km) may have amplitudes
smaller than those of body waves or may not even be detected on seismic records (see Figure
2 in EX 11.2). This should alert seismogram analysts to look for depth phases, which are then
usually well separated from their primary waves and so are easily recognized (see Fig. 11.7
above and Figure 6a and b in DS 11.2).
Another feature that helps in phase identification is the waveform. Most striking is the
difference in waveforms between body and surface waves. Dispersion in surface waves results
in long wave trains of slowly increasing and then decreasing amplitudes, whereas non-
dispersive body waves form short duration wavelets. Usually, the longer period waves arrive
first (“normal” or “positive” dispersion) (see Figs. 11.8 and 11.9). However, the very long-
period waves (T > 60 s) , that penetrate into the mantle down to the asthenosphere (a zone of
low wave speeds), may show inverse dispersion. The longest waves then arrive later in the
wave train (see Fig. 2.18).
For an earthquake of a given seismic moment, the maximum amplitude of the S wave is about
five-times larger at source than that of the P waves (see Figs. 2.3, 2.23, and 2.41). This is a
consequence of the different propagation velocities of P and S waves (see Eq. (3.2). Also the
spectrum is different for each wave type. Thus, P-wave source spectra have corner
frequencies about √3 times higher than those of S. In high-frequency filtered records this may
increase P-wave amplitudes with respect to S-wave amplitudes (see Fig. 11.10 right).
Additionally, the frequency-dependent attenuation of S waves is significantly larger than for P
waves.
P P
Fig. 11.10 Left: Low-pass filtered (< 0.1 Hz) and right: band-pass filtered (3.0-8.0 Hz)
seismograms of the Oct. 16, 1999, earthquake in California (mb = 6.6, Ms = 7.9) as recorded
at the broadband station DUG at D = 6° (courtesy of L. Ottemöller).
Due to both effects, S waves and their multiple reflections and conversions are – within the
teleseismic distance range – mainly observed on LP or BB records. On the other hand, the
different P-wave phases, such as P, PcP, PKP, and PKKP, are well recorded, up to the largest
epicentral distances, by SP seismographs with maximum magnification typically around 1 Hz.
Generally, the rupture duration of earthquakes is longer than the source process of explosions.
It ranges from less than a second for small microearthquakes up to several minutes for the
largest shallow crustal shocks with a source which is usually a complex multiple rupture
process (see Fig. 11.11, Fig. 3.7 and Figure 5b in DS 11.2).
Fig. 11.11 Vertical component records of the P-wave group from a crustal earthquake in
Sumatra (04 June 2000; mb = 6.8, Ms = 8.0) at the GRSN station MOX at D = 93.8°. Top:
WWSSN-SP (type A); middle: medium-period Kirnos SKD BB-displacement record (type
C), and bottom: original BB-velocity record. Clearly recognizable is the multiple rupture
process with P4 = Pmax arriving 25 s after the first arrival P1. The short-period magnitude mb
determined from P1 would be only 5.4, mb = 6.3 from P2 and mb = 6.9 when calculated from
P4. When determining the medium-period body-wave magnitude from P4 on the Kirnos
record then mB = 7.4.
As compared to shallow crustal earthquakes, deep earthquakes of comparable magnitude are

often associated with higher stress drop and smaller source dimension. This results in the
strong excitation of higher frequencies and thus simple and impulse-like waveforms (see Fig.
4.13 and Figures 6a and b in DS 11.2). Therefore S waves from deep earthquakes may be
recognizable in short-period records even at teleseismic distances. The same applies to
waveforms from explosions. As compared to shallow earthquakes, when scaled to the same
magnitude, their source dimension is usually smaller, their source process simpler and their
source duration much shorter (typically in the range of milliseconds). Accordingly, explosions
generate significantly more high-frequency energy than earthquakes and usually produce
shorter and simpler waveforms. Examples are given in Figures 1 to 5 of DS 11.4. Note,
however, that production explosions in large quarries or open cast mines, with yields ranging
from several hundred to more than one kiloton TNT, are usually fired in sequences of time-
delayed sub-explosions, which are spread out over a large area. Such explosions may generate
rather complex wave fields, waveforms and unusual spectra, sometimes further complicated
by the local geology and topography, and thus not easy to discriminate from local
earthquakes.
At some particular distances, body waves may have relatively large amplitudes, especially
near caustics (see Fig. 2.29 for P waves in the distance range between 15° and 30°; or around
D = 145° for PKP phases). In contrast, amplitudes decay rapidly in shadow zones (such as for
P waves beyond 100°; see Fig. 11.63). The double triplication of the P-wave travel-time curve
between 15° and 30° results in closely spaced successive onsets and consequently rather
complex waveforms (Fig. 11.49). At distances between about 30° and 100°, however,
waveforms of P may be simple (see Figs. 11.52 and 11.53). Beyond the PKP caustic, between
145° < D < 160°, longitudinal core phases split into three travel-time branches with typical
amplitude-distance patterns. This, together with their systematic relative travel-time
differences, permits rather reliable phase identification and distance estimates, often better
than 1° (see Figs. 11.62 and 11.63 as well as exercise EX 11.3).
Fig. 11.12 is a simplified diagram showing the relative frequency of later body-wave arrivals
with respect to the first arrival P or the number n of analyzed earthquakes, as a function of
epicentral distance D between 36° and 166°. They are based on observations in standard
records (see Fig. 3.11) of types A4 (SP - short-period, < 1.5 s), B3 (LP – long-period, between
20 s and 80 s) and C (BB - broadband displacement between 0.1 s and 20 s) at station MOX
in Germany (Bormann, 1972a). These diagrams show that in the teleseismic distance range
one can mainly expect to observe in SP records the following longitudinal phases: P, PcP,
ScP, PP, PKP (of branches ab, bc and df), P'P' (= PKPPKP), PKKP, PcPPKP, SKP and the
depth phases of P, PP and PKP. In LP and BB records, however, additionally S, ScS, SS,
SSS, SKS, SKSP, SKKS, SKKP, SKKKS, PS, PPS, SSP and their depth phases are
frequently recorded. This early finding based on the visual analysis of traditional analog film
recordings has recently been confirmed by stacking SP and LP filtered broadband records of
the Global Digital Seismic Network (GDSN) (Astiz et al., 1996; see Figs. 2.47 and 2.48 with
overlays).
Since these diagrams and stacked seismogram sections reflect, in a condensed form, some
systematic differences in waveforms, amplitudes, dominating periods and relative frequency
of occurrence of seismic waves in different distance ranges, they may, when used in addition
to travel-time curves, give some guidance to seismogram analysts as to what kind of phases
they may expect at which epicentral distances and in which kind of seismic records. Note,
however, that the appearance of these phases is not “obligatory”, rather, it may vary from
region to region, depending also on the source mechanisms and the radiation pattern with
respect to the recording station, the source depth, the area of reflection (e.g., underneath
oceans, continental shield regions, young mountain ranges), and the distance of the given
station from zones with frequent deep earthquakes. Therefore, no rigid rules for phase
identification can be given. Also, Fig. 11.12 considers only teleseismic earthquakes. Local
and regional earthquakes, however, are mainly recorded by SP short-period seismographs of
type A or with Wood-Anderson response. There are several reasons for this. Firstly, SP
seismographs have usually the largest amplification and so are able to record (at distances
smaller than a few hundred kilometers) sources with magnitudes of zero or even less.
Secondly, as follows from Fig. 3.5, the corner frequency of source displacement spectra for
events with magnitudes < 4 is usually > 1 Hz, i.e., small events radiate relatively more high-
frequency energy. Thirdly, in the near range the high frequencies have not yet been reduced
so much by attenuation and scattering, as they usually are for f > 1 Hz in the teleseismic
range. Therefore, most local recordings show no waves with periods longer than 2 s.
However, as Ml increases above 4, more and more long-period waves with large amplitudes
are generated and these dominate in BB records of local events, as illustrated with the records
in Figs. 11.8 and 11.10.
Fig. 11.12 Relative frequency of occurrence of secondary phases in standard analog records
at station MOX, Germany, within the teleseismic distance range 36° to 166°. The first column
relates to 100% of analyzed P-wave first arrivals or of analyzed events (hatched column),
respectively. In the boxes beneath the phase columns the type of standard records is indicated
in which these phases have been observed best or less frequently/clear (then record symbols
in brackets). A – short-period; B – long-period LP, C – Kirnos SKD BB-displacement.
11.2.4.3 Polarization
As outlined in 2.2 and 2.3, P and S waves are linearly polarized, with slight deviations from
this ideal in the inhomogeneous and partially anisotropic real Earth (see Figs. 2.6 and 2.7). In
contrast, surface waves may either be linearly polarized in the horizontal plane perpendicular
to the direction of wave propagation (transverse polarization; T direction; e.g., Love waves)
or elliptically polarized in the vertical plane oriented in the radial (R) direction of wave
propagation (see Figs. 2.8, 2.13 and 2.14). P-wave particle motion is dominatingly back and
forth, parallel to the seismic ray, whereas S-wave motion is perpendicular to the ray direction.
Accordingly, a P-wave motion can be split into two main components, one vertical (Z) and
one horizontal (R) component. The same applies to Rayleigh waves, but with a 90° phase
shift between the Z and R components of motion. S waves, on the other hand, may show
purely transverse motion, oscillating in the horizontal plane (SH; i.e., pure T component, as
Love waves) or motion in the vertical propagation plane, at right angles to the ray direction
(SV), or in any other combination of SH and SV. In the latter case S-wave particle motion has
Z, R and T components, with SV wave split into a Z and an R component.
Thus, when 3-component records are available, the particle motion of seismic waves in space
can be reconstructed and used for the identification of seismic wave types. However, usually
the horizontal seismometers are oriented in geographic east (E) and north (N) direction. Then,
first the backazimuth of the source has to be computed (see EX 11.2) and then the horizontal
components have to be rotated into the horizontal R direction and the perpendicular T
direction, respectively. This axis rotation is easily performed when digital 3-component data
and suitable analysis software are available. It may even be carried one step further by
rotating the R component once more into the direction of the incident seismic ray
(longitudinal L direction). The T component then remains unchanged but the Z component is
rotated into the Q direction of the SV component. Such a ray-oriented co-ordinate system
separates and plots P, SH and SV waves in 3 different components L, T and Q, respectively.
These axes transformations are easily made given digital data from arbitrarily oriented
orthogonal 3-component sensors such as the widely used triaxial sensors STS2 (see Fig. 5.13
and DS 5.1). However, the principle types of polarization can often be quickly assessed with
manual measurement and elementary calculation from analog 3-component records and the
backazimuth from the station to the source be estimated (see EX 11.2).
Note that all direct, reflected and refracted P waves and their multiples, as well as conversions
from P to S and vice versa, have their dominant motion confined to the Z and R (or L and Q)
plane. This applies to all core phases, also to SKS and its multiples, because K stands for a P-
wave leg in the outer core. In contrast, S waves may have both SV and SH energy, depending
on the source type and rupture orientation. However, discontinuities along the propagation
path of S waves act as selective SV/SH filters. Therefore, when an S wave arrives at the free
surface, part of its SV energy may be converted into P, thus forming an SP phase.
Consequently, the energy reflected as S has a larger SH component as compared to the
incoming S. So the more often a mixed SH/SV type of S wave is reflected at the surface, the
more it becomes of SH type. Accordingly, SSS, SSSS etc. will show up most clearly or even
exclusively on the T component (e.g., Fig. 11.37) unless the primary S wave is dominantly of
SV-type (e.g., Fig. 11.13). As a matter of fact, Love waves are formed through constructive
interference of repeated reflections of SH at the free surface. Similarly, when an S wave hits
the core-mantle boundary, part of its SV energy is converted into P which is either refracted
into the core (as K) or reflected back into the mantle as P, thus forming the ScP phase.
Consequently, multiple ScS is also usually best developed on the T component.
Fig. 11.13 shows an example of the good separation of several main seismic phases on an Z-
R-T-component plot. At such a large epicentral distance (D = 86.5°) the incidence angle of P
is small (about 15°; see EX 3.3). Therefore, the P-wave amplitude is largest on the Z
component whereas for PP, which has a significantly larger incidence angle, the amplitude on
the R component is almost as large as Z. For both P and PP no T component is recognizable
above the noise. SKS is strong in R and has only a small T component (effect of anisotropy,
see Fig. 2.7). The phase SP has both a strong Z and R component. Love waves (LQ) appear as
the first surface waves in T with very small amplitudes in R and Z. In contrast, Rayleigh
waves (LR) are strongest in R and Z. SS in this example is also largest in R. From this one
can conclude, that the S waves generated by this earthquake are almost purely of SV type. In
other cases, however, it is only the difference in the R-T polarization which allows S to be
distinguished from SKS in this distance range of around 80° where these two phases arrive
closely to each other (see Fig. 11.14 and Figure 13e in DS 11.2).
Fig. 11.13 Time-compressed long-period filtered three-component seismogram (SRO-LP

simulation filter) of the Nicaragua earthquake recorded at station MOX (D = 86.5°).
Horizontal components have been rotated (ZRT) with R (radial component) in source
direction. The seismogram shows long-period phases P, PP, SKS, SP, SS and surface waves L
(or LQ for Love wave) and R (or LR for Rayleigh wave).
Fig. 11.14 Ray-oriented broadband records ( left: Z-N-E components; right: particle motion
in the Q-T plane) of the S and SKS wave group from a Hokkaido Ms = 6.5 earthquake on 21
March 1982, at station Kasperske Hory (KHC) at an epicentral distance of D = 78.5°.
The empirical travel-time curves in Fig. 2.49 (from Astiz et al., 1996) summarize rather well,
which phases (according to the overlay of Fig. 2.48) are expected to dominate the vertical,
radial or transverse ground motion in rotated three-component records.
If we supplement the use of travel-time curves with seismic recordings in different frequency
bands, and take into account systematic differences in amplitude, frequency content and
polarization for P, S and surface waves, and when we know the distances, where caustics and
shadow zones occur, then the identification of later seismic wave arrivals is entertaining and
like a detective inquiry into the seismic record.
11.2.4.4 Example for documenting and reporting of seismogram parameter readings
Fig. 11.15 shows a plot of the early part of a teleseismic earthquake recorded at stations of the
GRSN. At all stations the first arriving P wave is clearly recognizable although the P-wave
amplitudes vary strongly throughout the network. This is not a distance effect (the network
aperture is less than 10% of the epicentral distance) but rather an effect of different local site
conditions related to underground geology and crustal heterogeneity. As demonstrated with
Figs. 4.35 and 4.36, the effect is not a constant for each station but depends both on azimuth
and distance of the source. It is important to document this. Also, Fig. 11.15 shows for most
stations a clear later arrival about 12 s after P. For the given epicentral distance, no other main
phase such as PP, PPP or PcP can occur at such a time (see differential travel-time curves in
Figure 4 of EX 11.2). It is important to pick such later (so-called secondary) onsets which
might be “depth phases” (see 11.2.5.1) as these allow a much better determination of source
depth than from P-wave first arrivals alone (see Figure 7 in IS 11.1).
Fig. 11.15 WWSSN-SP filtered seismograms at 14 GRSN, GRF, GERESS and GEOFON
stations from an earthquake in Mongolia (24 Sept. 1998; depth (NEIC-QED) = 33 km; mb =
5.3, Ms = 5.4). Coherent traces have been time-shifted, aligned and sorted according to
epicentral distance (D = 58.3° to BRG, 60.4° to GRA1 and 63.0° to WLF). Note the strong
variation in P-wave signal amplitudes and clear depth phases pP arriving about 12 s after P.
Tab. 11.2 gives for the Mongolia earthquake shown in Fig. 11.15 the whole set of parameter
readings made at the analysis center of the Central Seismological Observatory Gräfenberg
(SZGRF) in Erlangen, Germany:
• first line: date, event identifier, analyst;

• second and following lines: station, onset time, onset character (e or i), phase name
(P, S, etc.), direction of first particle motion (c or d), analyzed component, period
[s], amplitude [nm], magnitude (mb or Ms), epicentral distance [°]; and
• last two lines: source parameters as determined by the SZGRF (origin time OT,
epicentre, average values of mb and Ms, source depth and name of Flinn-Engdahl-
region).
Generally, these parameters are stored in a database, used for data exchange and published in
lists, bulletins and the Internet (see IS 11.2). The onset characters i (impulsive) should be used
only if the time accuracy is better than a few tens of a second, otherwise the onset will be
described as e (emergent). Also, when the signal-to-noise-ratio (SNR) of onsets is small and
especially, when narrow-band filters are used, the first particle motion should not be given
because it might be distorted or lost in the noise. Broadband records are better suited for
polarity readings (see Fig. 4.10). Their polarities, however, should be reported as u (for “up”
= compression) and r (for “rarefaction” = dilatation) so as to differentiate them from short-
period polarity readings (c and d, respectively).
Tab. 11.2 Parameter readings at the SZGRF analysis center for the Mongolia earthquake
shown in Fig. 11.15 from records of the GRSN.
Note that for this event the international data center NEIC had “set” the source depth to 33 km
because of the absence of reported depth phases. The depth-phase picks at the GRSN,
however, with an average time difference of pP-P of about 12 s, give a focal depth of 44 km.
Also note in Tab. 11.2 the large differences in amplitudes (A) determined from the records of
individual stations. The resulting magnitudes mb vary between 5.4 (GEC2) and 6.2 (GRA1)!
11.2.5 Criteria to be used in event identification and discrimination
11.2.5.1 Discrimination between shallow and deep earthquakes
Earthquakes are often classified on depth as: shallow focus (depth between 0 and 70 km),
intermediate focus (depth between 70 and 300 km) and deep focus (depth between 300 and
700 km). However, the term "deep-focus earthquakes" is also often applied to all sub-crustal
earthquakes deeper than 70 km. They are generally located in slabs of the lithosphere which
are subducted into the mantle. As noted above, the most obvious indication on a seismogram
that a large earthquake has a deep focus is the small amplitude of the surface waves with
respect to the body-wave amplitudes and the rather simple character of the P and S
waveforms, which often have impulsive onsets (see Fig. 4.13). In contrast to shallow-focus
earthquakes, S phases from deep earthquakes may sometimes be recognizable even in
teleseismic short-period records. The body-wave/surface-wave ratio and the type of generated
surface waves are also key criteria for discriminating between natural earthquakes, which
mostly occur at depth larger than 5 km, and quarry blasts, underground explosions or
rockbursts in mines, which occur at shallower depth (see 11.2.5.2).
A more precise determination of the depth h of a seismic source, however, requires either the
availability of a seismic network with at least one station being very near to the source, e.g., at
an epicentral distance D < h (because only in the near range the travel time t(D, h) of the
direct P wave varies strongly with source depth h), or the identification of seismic depth
phases on the seismic record. The most accurate method of determining the focal depth of an
earthquake in routine seismogram analysis, particularly when only single station or network
records at teleseismic distances are available, is to identify and read the onset times of depth
phases. A depth phase is a characteristic phase of a wave reflected from the surface of the
Earth at a point relatively near the hypocenter (see Fig. 2.43). At distant seismograph stations,
the depth phases pP or sP follow the direct P wave by a time interval that changes only
slowly with distance but rapidly with depth. The time difference between P and other primary
seismic phases, however, such as PcP, PP, S, SS etc. changes much more with distance. When
records of stations at different distances are available, the different travel-time behavior of
primary and depth phases makes it easier to recognize and identify such phases. Because of
the more or less fixed ratio between the velocities of P and S waves with vP/vs ≈ √3, pP and sP
follow P with a more or less fixed ratio of travel-time difference t(sP-P) ≈ 1.5 t(pP-P) (see
Figs. 11.16 and 11.17). Animations of seismic ray propagation and phase recordings from
deep earthquakes are given in files 3 and 5 of IS 11.3 and related CD-ROM.
The time difference between pP and sP and other direct or multiple reflected P waves such as
pPP, sPP, pPKP, sPKP, pPdif, sPdif, etc. are all roughly the same. S waves also generate
depth phases, e.g., sS, sSKS, sSP etc. The time difference sS-S is only slightly larger than sP-
P (see Figs. 1.4 and 11.17). The difference grows with distance to a maximum of 1.2 times the
sP-P time. These additional depth phases may also be well recorded and can be used in a
similar way for depth determination as pP and sP.
Given the rough distance between the epicenter and the station, the hypocenter depth (h) can
be estimated within ∆h ≈ ±10 km from travel-time curves or determined by using time-
difference tables for depth-phases (e.g., from ∆t(pP-P) or ∆t(sP-P); see Kennett, 1991 or
Table 1 in EX 11.2) or the “rule-of-thumb” in Eq. (11.4). An example is given in Fig. 11.18.
It depicts broadband records of the GRSN from a deep earthquake (h = 119 km) in the
Volcano Islands, West Pacific. The distance range is 93° to 99°. The depth phases pP and pPP
are marked. From the time difference pP-P of 31.5 s and an average distance of 96°, it follows
from Table 1 in EX 11.2 that the source depth is 122 km. When using Eq. (11.4) instead, we
get h = 120 km. This is very close to the source depth of h = 119 km determined by NEIC
from data of the global network.
Note that on the records in Fig. 11.18 the depth phases pP and pPP have larger amplitudes
than the primary P wave. This may be the case also for sP, sS etc., if the given source
mechanism radiates more energy in the direction of the upgoing rays (p or s; see Fig. 2.43)
than in the direction of the downgoing rays for the related primary phases P, PP or S. Also, in
Fig. 11.18, pP, PP and pPP have also longer periods than P. Accordingly, they are more
coherent throughout the network than the shorter P waves. Fig. 11.37 shows for the same
earthquake the LP-filtered and rotated 3-component record at station RUE, Germany, with all
identified major later arrivals being marked on the record traces. This figure is an example of
the search for and comprehensive analysis of secondary phases.
Fig. 11.16 Short-period (left) and long-period (right) seismograms from a deep-focus Peru-
Brazil border region earthquake on May 1, 1986 (mb = 6.0, h = 600 km) recorded by stations
in the distance range 50.1° to 92.2°. Note that the travel-time difference between P and its
depth phases pP and sP, respectively, remains nearly unchanged. In contrast PcP comes closer
to P with increasing distance and after merging with P at joint grazing incidence on the core-
mantle boundary form the diffracted wave Pdif (reprinted from Anatomy of Seismograms,
Kulhánek, Plate 41, p. 139-140;  1990; with permission from Elsevier Science).
Fig. 11.17 3-component recordings in the distance range 18.8° to 24.1° from a regional
network of portable BB instruments deployed in Queensland, Australia (seismometers
CMG3ESP; unfiltered velocity response; see DS 5.1). The event occurred in the New
Hebrides at 152 km depth. On each set of records the predicted phase arrival times for the
AK135 model (see Fig. 2.53) are shown as faint lines. The depth phases pP, sP and sS are
well developed but their waveforms are complex because several of the arrivals have almost
the same travel time (courtesy of B. Kennett).
Crustal earthquakes usually have a source depth of less than 30 km, so the depth phases may
follow their primary phases so closely that their waveforms overlap (see Fig. 11.19).
Identification and onset-time picking of depth phases is then usually no longer possible by
simple visual inspection of the record. Therefore, in the absence of depth phases reported by
seismic stations, international data centers such as NEIC in its Monthly Listings of
Preliminary (or Quick) Determination of Epicenters often fix the source depth of (presumed)
crustal events at 0 km, 10 km or 33 km, as has been the case for the event shown in Fig.
11.15. This is often further specified by adding the capital letter N (for “normal depth” = 33
km) of G (for depth fixed by a geophysicist/analyst). Waveform modeling, however (see 2.8
and Figs. 2.57 to 2.59), may enable good depth estimates for shallow earthquakes to be
obtained from the best fit of the observed waveforms to synthetic waveforms calculated for
different source depth. Although this is not yet routine practice at individual stations, the
NEIC has, since 1996, supplemented depth determinations from pP-P and sP-P by synthetic
modeling of BB-seismograms. The depth determination is done simultaneously with the
determination of fault-plane solutions. This has reduced significantly the number of

earthquakes in the PDE listings with arbitrarily assigned source depth 10G or 33N.
Fig. 11.18 Broadband vertical-component seismograms of a deep (h = 119 km) earthquake

from Volcano Islands region recorded at 17 GRSN, GRF and GEOFON stations. (Source
data by NEIC: 2000-03-28 OT 11:00:21.7 UT; 22.362°N, 143.680°E; depth 119 km; mb 6.8;
D = 96.8° and BAZ = 43.5° from GRA1). Traces are sorted according to distance.
Amplitudes of P are smaller than pP. Phases with longer periods PP, pP and pPP are much
more coherent than P.
Note, however, that often there is no clear evidence of near-source surface reflections in
seismic records, or they show apparent pP and sP but with times that are inconsistent from
station to station. Douglas et al. (1974 and 1984) have looked into these complexities,
particularly in short-period records. Some of these difficulties are avoided in BB and LP
recordings. Also, for shallow sources, surface-wave spectra may give the best indication of
depth but this method is not easy to apply routinely. In summary, observational seismologists
should be aware that depth phases are vital for improving source locations and making
progress in understanding earthquakes in relation to the rheological properties and stress
conditions in the lithosphere and upper mantle. Therefore, they should do their utmost to
recognize depth phases in seismograms despite the fact that they are not always present and
that it may be difficult to identify them reliably.
More examples of different kinds of depth phases are given in Figs. 11.34 and 11.35d as well
as in Figure 6b of DS 11.2 and Figures 1b, 2b, 5b and 7a +b in DS 11.3.
Fig. 11.19 3-component records in the distance range between 7.9° and 21.1° by a regional
network of portable broadband instruments deployed in Queensland, Australia (seismometers
CMG3ESP; unfiltered velocity response). The event occurred in Papua New Guinea at 15 km
depth. As in Fig. 11.17 the predicted phase arrival times for the AK135 model are depicted.
Primary, depth and other secondary arrivals (such as PnPn in the P-wave group and SbSb as
well as SgSg in the S-wave group) superpose to complex wavelets. Also note that several of
the theoretically expected phases have such weak energy that they can not be recognized on
the records at the marked predicted arrival times above the noise level or the signal level of
other phases (e.g., PcP at most stations) (courtesy of B. Kennett).
11.2.5.2 Discrimination between natural earthquakes and man-made seismic events

Quarry and mining blasts, besides dedicated explosion charges in controlled-sources
seismology, may excite strong seismic waves. The largest of these events may have local
magnitudes in the range 2 to 4 and may be recorded over distances of several hundred
kilometers. Rock bursts or collapses of large open galleries in underground mines may also
generate seismic waves (see Figure 3 in EX 11.1). The magnitude of these induced seismic
events may range from around 2 to 5.5 and their waves may be recorded world-wide (as it
was the case with the mining collapse shown in Fig. 11.8). In some countries with low to
moderate natural seismicity but a lot of blasting and mining, anthropogenic (so-called “man-
made” or “man-induced”) events may form a major fraction of all recorded seismic sources,
3and may even outnumber recordings of earthquakes. Then a major seismological challenge
is the reliable discrimination of different source types. Fig. 11. 39 shows a comparison of
seismograms from: (a) a mining-induced earthquake; (b) a quarry blast; (c) a local earthquake;
(d) a regional earthquake; and (e) a teleseismic earthquake. Seismograms (a) and (b) show
that the high-frequency body-wave arrivals are followed, after Sg, by well developed lower-
frequency and clearly dispersed Rayleigh surface waves (Rg; strong vertical components).
This is not so for the two earthquake records (c) and (d) because sources more than a few
kilometers deep do not generate short-period fundamental Rayleigh waves of Rg type. For
even deeper (sub-crustal) earthquakes (e.g., Fig. 2.41) only the two high-frequency P- and S-
wave phases are recorded within a few hundred kilometers from the epicenter.
Based on these systematic differences in frequency content and polarization, some

observatories that record many quarry blasts and mining events, such as GRFO, have
developed automatic discrimination filters to separate them routinely from tectonic
earthquakes. Chernobay and Gabsatarova (1999) give references to many other algorithms
for (semi-) automatic source classification. These authors tested the efficiency of the
spectrogram and the Pg/Lg spectral ratio method for routine discrimination between regional
earthquakes with magnitudes smaller than 4.5 and chemical (quarry) explosions of
comparable magnitudes based on digital records obtained by a seismic network in the
Northern Caucasus area of Russia. They showed that no single method can yet assure reliable
discrimination between seismic signals from earthquakes and explosions in this region.
However, by applying a self-training algorithm, based on hierarchical multi-parameter cluster
analysis, almost 98% of the investigated events could be correctly classified and separated
into 19 groups of different sources. However, local geology and topography as well as
earthquake source mechanisms and applied explosion technologies may vary significantly
from region to region (see page 18 of this Chapter). Therefore, there exists no straightforward
and globally applicable set of criteria for reliable discrimination between man-made and
natural earthquakes.
In this context one should also discuss the discrimination between natural earthquakes (EQ)
and underground nuclear explosions (UNE). The Comprehensive Nuclear-Test-Ban Treaty
(CTBT) has been negotiated for decades as a matter of high political priority. A Preparatory
Commission for the CTBT Organization (CTBTO) has been established with its headquarters
in Vienna (http://www.ctbto.org) which is operating an International Monitoring System
(IMS; see http://www.nemre.nn.doe.gov/nemre/introduction/ims_descript.html, Fig. 8.12 and
Barrientos et al., 2001). In the framework of the CTBTO, initially a Prototype International
Data Centre (PIDC; http://www.pidc.org/) was established in Arlington, USA, which is
replaced since 2001 by the International Data Centre in Vienna. Agreement was reached only
after many years of demonstrating the potential of seismic methods to discriminate
underground explosions from earthquakes, down to rather small magnitudes mb ≈ 3.5 to 4.
Thus, by complementing seismic event detection and monitoring with hydroacoustic,
infrasound and radionuclide measurements it is now highly probable that test ban violations
can be detected and verified.
The source process of UNEs is simpler and much shorter than for earthquake shear ruptures
(see Figs. 3.3 – 3.5 and related discussions). Accordingly, P waves from explosions have
higher predominant frequencies and are more like impulses than earthquakes and have
compressional first motions in all directions. Also, UNEs generate lower amplitude S and
surface waves than earthquakes of the same body-wave magnitude (see Fig. 11.20).
Fig. 11.20 Broadband displacement records of an earthquake and an underground nuclear
explosion (UNE) of comparable magnitude and at nearly the same distance (about 40°).
Fig. 11.21 Short-period records at station MOX a) of an underground nuclear explosion at

the Semipalatinsk (SPT) test site in Kazakhstan (D = 41°) and b) of an earthquake with
comparable magnitude and at similar distance.
In short-period records of higher time resolution the difference in frequency content,

complexity and duration of the P-wave group between underground nuclear explosions and
earthquakes is often clear. Fig. 11.21 gives an example. As early as 1971 Weichert developed
an advanced short-period spectral criterion for discriminating between earthquakes and
explosions and Bormann (1972c) combined in a single complexity factor K differences in
frequency content, signal complexity and duration to a powerful heuristic discriminant.
Another powerful discriminant is the ratio between short-period P-wave magnitude mb and
long-period surface-wave magnitude Ms. The former samples energy around 1 Hz while the
latter samples long-period energy around 0.05 Hz. Accordingly, much smaller Ms/mb ratios
are observed for explosions than for earthquakes (see Fig. 11.20). Whereas for a global
sample of EQs and UNEs the two population overlap in an Ms/mb diagram, the separation is
good when earthquakes and explosions in the same region are considered (Bormann, 1972c).
Early studies have shown that with data of mb ≥ 5 from only one teleseismic station 100% of
the observed UNEs with magnitudes from the SPT test site could be separated from 95% of
the EQs in Middle Asia, whereas for the more distant test site in Nevada (D = 81°) 95% of the
UNEs could be discriminated from 90% of the EQs in the Western USA and Middle America
(see Fig. 11.22).
Fig. 11.22 Separation of EQs and UNEs by the Ms/mb criterion according to data collected at
station MOX, Germany. Left: for Middle Asia and the test site in Semipalatinsk (SPT), right:
for USA/Middle America and the Nevada test site (NTS).
Other potential discrimination criteria, such as the different azimuthal distribution of P-wave
first-motion polarities expected from UNEs (always +) and EQs (mixed + and -), have not
proved to be reliable. One reason is, that due to the narrowband filtering, which is applied to
reach the lowest possible detection threshold, the P waveform, and particularly the first half
cycle, is often so much distorted, that the real first-motion polarity is no longer recognizable
in the presence of noise (see Fig. 4.10). Detailed investigations also revealed that simplified
initial model assumptions about the difference between explosion and earthquake sources do
not hold true. Surprisingly, the explosion source is poorly understood and source dimensions
around magnitude mb seem to be the same for earthquakes and explosions. Also, many
explosions do not approximate to a point-like expansion source in a half-space: significant
Love waves are generated (e.g., by Novaya Zemlya tests) and many P seismograms show
arrivals that can not be explained (see, e.g., Douglas and Rivers, 1988). Further, it has become
clear that much of the differences observed between records of UNEs and EQs are not due to
source differences but rather to differences in the geology, topography and seismotectonics of
the wider area around the test sites, and that this necessitates the calibration of individual
regions (e.g., Douglas et al., 1974).
In summary, one can say that the key criteria to separate EQs and explosions usually work
well for large events, however, difficulties come with trying to identify every EQ down to
magnitudes around mb = 4 with about 8000 earthquakes of this size per year. It is beyond the
scope of this section to go into more detail on this issue. Rather, the Editor has invited experts
from the CTBTO community to write for Volume 2 of the Manual a complementary
information sheet on advanced event detection and discrimination routines. This still
forthcoming information sheet will catalog the most important criteria, which have been
developed so far for discrimination and show more examples about their application to and
efficiency in different regions.
11.2.6 Quick event identification and location by means of single-station
three-component recordings
11.2.6.1 What is the best way of analyzing three-component seismograms?
Increasingly seismograms are being analyzed at laboratories that receive the data in (near) real
time from networks or arrays of seismometers (see Chapters 8 and 9). The seismograms can
then be analyzed jointly. Nevertheless, there remain many single, autonomous stations
operating around the world, in countries of the former Soviet Union and developing countries
in particular. Some of these single stations still record only with analog techniques. Yet much
can be done even under these “old-fashioned” conditions by the station personnel, provided
that at least some form of 3-component recording, either BB or both SP and LP, is available.
With such recordings it will be possible to assess quickly the source type, estimate its rough
location and magnitude, and identify in some detail later seismic phases, without waiting for
rapid epicenter determinations by international data centers before record analysis can begin.
Rather, there would be advantages if, in future, readings of secondary phases, particularly
depth phases, were reported as early as possible to regional and global data centers. Such
readings are indispensable for more accurate hypocenter location (see Figure 7 in IS 11.1).
Only recently both NEIC and the ISC began considering the introduction of more flexible and
sophisticated algorithms that can best make use of secondary phase readings for more reliable
(and rapid) hypocenter locations.
It has also been realized that accurate epicentral distances estimated from three-component
broadband readings of secondary phases can significantly improve location estimates by array
stations based purely on measurements of the P-wave vector slowness (originally by using
solely vertical-component SP sensors). Now, since modern software for digital seismogram
analysis has made it much simpler and faster than in the “analog past” to evaluate three-
component broadband data, we focus on such data here. Other procedures of modern multi-
station (but usually single-component) data analysis are dealt with later en passant. Array
analysis is discussed in detail in Chapter 9.
How then to proceed best in analyzing analog presentations of seismograms? The most
important rules, taking the discussion under 11.2.4 and 11.2.5 into account, are:
Take interest! Be curious! Ask questions!

to your seismic record
1. Are you NEAR (D < 20°) or TELESEISMIC (D > 20°)?

Criteria:
• Frequencies on SP records f ≥ 1 Hz f ≤ 1 Hz
• Amplitudes on LP records not or weaker large, also for later phases
• Record duration < 20 min > 20 min
(for magnitudes < 5; may be longer for strong earthquakes; see Fig. 1.2)
2. Is your D < 100° or D > 100° ?

Criteria:
• Surface wave max. after P arrival < 45 ± 5 min or > 45 ± 5 min (Table 5 in DS 3.1)
• Record duration on LP records < 1.5 hours or > 1.5 hours
(may be larger for very strong earthquakes; see Fig. 1.2)
3. Are you SHALLOW or DEEP (> 70 km)?

Criteria:
• Surface waves on LP records strong weak or none
• Depth phases usually not clear well separated and often clear
• Waveforms usually more complex more impulsive
4. Is the first strong horizontal arrival S or SKS ?

Criteria:
• Time difference to P < 10 ± 0.5 min ≈ 10 ± 0.5 min
• Polarization large horiz. A in R and/or T in R only
Warning ! If the first strong horizontal arrival follows P after ≈ 10 ± 0.5 min it may be SKS.
Check polarization! (see Fig. 11.14). Misinterpreting SKS as S may yield D estimates up to
20° too short. Look also for later multiple S arrivals (SP, SS, SSS) with better D control.
5. What are the first longitudinal and transverse onsets for D > 100° ?
Beyond 100° epicentral distance first arrivals may still be P, which may be seen particularly
in LP records of large earthquakes up to about almost 150° (see Fig. 11.63). This P, however,
has been diffracted around the core-mantle boundary and is termed Pdif (old Pdiff; see Figs.
11.59 and 11.63). First onsets in SP records are usually PKiKP and PKPdf (see Fig. 11.59),
or, somewhat later PP, which is often the first strong longitudinal Z-component arrival in both
SP, LP and BB records (see Figs. 11.60 and 11.63). The first strong arrivals on horizontal (R)
components are PKS or SKS. Misinterpretation of the first P-wave and S-wave arrivals as
direct P and S, respectively, may result in epicentral distance estimates up to more than 70°
too short! This can be avoided by taking the criteria under 2. into account. Also note, that the
travel-time difference between PKPdf and PKS or SKS is (almost) independent of distance.
The first arriving P and S waves do not then allow distance to be estimated. Therefore look
for later arriving multiple reflected S waves such as SS, SSS, etc., which are usually well
developed in this distance range on horizontal LP records and so allow D to to be estimated
with an error of usually < 2°.
Additionally, one might look for criteria discussed in sub-section 11.2.5.2 for discriminating
between explosions and earthquakes.
If only very broadband digital records are available, which are usually proportional to ground
velocity, it is best to filter them to produce standard analog WWSSN-SP and -LP
seismograms before starting a reconnaissance analysis. One may also simulate Kirnos SKD
BB-displacement and Wood-Anderson (WA) SP-displacement seismograms (for response
characteristics see Fig. 3.11), because all these responses are required for proper magnitude
estimation according to established standards. Only after these seismograms have been
produced should one begin with the detailed analysis. The analysis might include phase
identification, picking of onset times, amplitudes and periods, and the application, if required,
of special filters, such as the ones for inverse Hilbert transformation of phases, which have
been distorted by traveling through internal caustics (see 2.5.4.3), or for separating phases on
their polarization to improve phase discrimination.
Of course, in countries with many seismic sources recorded every day it will not be possible,
particularly for untrained interpreters, to apply all these criteria to every seismic signal. On
the other hand, this kind of checking takes only a few seconds, minutes at most, for an
experienced interpreter who has already trained himself/herself in recognizing immediately
the different record patterns on seismograms from systems with standard responses. In
addition, many data centers specialize in analyzing only seismograms from local, regional or
teleseismic sources. Accordingly, either the number of questions to be asked to the record or
the number of signals to be analyzed will be reduced significantly. Also, the task might be
significantly eased at observatories or analysis centers which have advanced routines
available for digital seismogram analysis such as SEISAN or Seismic Handler. Provided that
first hypocenter estimates are already available from international data centers or from
analysis of array or network recordings, these computer programs allow the theoretical onset
times of expected seismic phases to be displayed on the seismogram. However, these
theoretical times should not be followed blindly but considered only as assistance. The
additional information on amplitudes, frequency content, and polarization has to be taken into
account before giving a name to a recognizable onset! (see 11.2.4 and 11.2.5).
On the other hand, it is meaningless to list more detailed and strict criteria and rules about the
appearance and identification of seismic phases in different distance ranges, because they vary
from event to event and from source region to source region. They also depend on the specific
conditions of the given propagation paths and the local environment at the receiving station.
Therefore, every station operator or network analyst has to develop, through experience and
systematic data analysis, his/her own criteria for improved seismogram analysis, source
identification, and location. In any event, however, the general approach to record analysis
given above should be followed to avoid the analysis becoming thoughtless, boring and
routine, which will inevitably result in the reporting of inhomogeneous and incomplete low-
quality data of little value for research or to the general user.
11.2.6.2 Hypocenter location
If well calibrated 3-component broadband and/or long-period recordings are available then it
is possible to locate sufficiently strong local events (Ml > 3) and teleseismic sources (mb > 5)
with an accuracy comparable to or even better than those for un-calibrated arrays or station
networks. This was demonstrated more than 30 years ago (Bormann, 1971a and b) by using
standard film records of type A, B and C (responses see Fig. 3.11). Amplitudes and onset
times were at that time still measured by using an ordinary ruler or a sub-millimeter scaled
magnification lens. Nevertheless, the mean square error of epicenters thus located within the
distance range 20° < D < 145° was less than 300 km when compared with the epicenter
coordinates published by the seismological World Data Centers A and B. Fig. 11.23 shows
the statistical distribution of errors in azimuth and distance based on several hundred 3-
component event locations.
Note that the errors in distance estimated from readings of P and later secondary phases
within the distance range 80° < D < 120° are mostly less than about ±1° and rarely greater
than ±2° . The mean errors seldom differ significantly from zero, and where they do it is
usually for specific regions (distance/azimuth ranges). Taking such systematic errors into
account, the location accuracy can be improved. Many seismic arrays and networks now use
routinely multi-phase epicentral distance determinations for improving their slowness-based
source locations. Some advanced software for seismogram analysis like SHM (see 11.4.1)
includes this complementary interactive analysis feature.
Backazimuth derived from SP 3-component recordings may have large systematic errors up to
several tens of degrees. This is not so if LP or BB records are used. Whereas individual
determinations of backazimuth from SP records might deviate up to about 40° from the true
source azimuth, the errors are rarely (except at low SNR) larger than 10° when BAZ is
determined from BB records (provided that the magnification of the horizontal components is
known with high accuracy or identical!). The reason for this is obvious from Fig. 2.6 and the
related discussion. The particle motion in SP records is complicated and random due to wave
scattering and diffraction by small-scale heterogeneities in the crust and by rough surface
topography at or near the station site (see Buchbinder and Haddon, 1990). In contrast, LP or
BB records, which are dominated by longer wavelength signals, usually show simpler P
waveforms with clearer first-motion polarity than do SP records. In addition, later phase
arrivals, which are crucial for accurate distance determination from single station records,
stand out more clearly or are recognizable only in BB or LP records (see Fig. 11.24).
Fig. 11.23 Left: Errors in backazimuth Az (or BAZ) at station MOX estimated using 3-
component records of type A (SP) and of type C (Kirnos SKD BB-displacement). Right:
Errors in estimating the epicentral distance D at station MOX from records of type C using
travel-time difference S-P in the distance range 10° < D < 100° or travel-time differences
between other seismic phases for D > 100°. The solid lines give the 90% confidence interval
for the mean error with number of the observations; the dash-dot lines are the 90%
confidence interval for a single observation.
Simple 3-component event locations based solely on readings of onset times of identified
phases, polarity of P-wave first motions and horizontal component amplitude ratio should
proceed as follows:
• general event classification (near/far; shallow/deep; D < 100°/> 100° etc.);

• picking and identifying the most pronounced phases by comparing the 3-component
record traces and related polarization characteristics (Fig. 11.24);
• determination of D by a) matching the identified body-wave phases with either
overlays of differential travel-time curves of equal time scale (see Figures 2 to 4 in
EX 11.2), b) by measuring their onset-time differences and comparing them with
respective distance-dependent differential travel-time tables or c) by computer
calculation of D based on digital time picks for identified phases and local, regional
and/or global travel-time models integrated into the analysis program;
• determination of source depth h on the basis of identified depth phases (see
11.2.5.1) and following correction of D, again by using either travel-time curves,
differential t-D tables or computer assisted time-picks and comparison with travel-
time models;
• determination of the backazimuth (against North) from the station to the source
from the first-motion directions in the original Z, N and E component records and
from the amplitude ratio AE/AN. For details see Figure 1 and explanations given in
EX 11.2;
• determination of the epicenter location and coordinates by using appropriate map
projections with isolines of equal azimuth and distance from the station (see Figure 5 in EX
11.2) or by means of suitable computer map projections.
Fig. 11.24 Left: Low-pass filtered digital broadband record of the Global Seismograph
Network (GSN) station KIV from the shallow (h = 10 km) Greece earthquake of 07 Sept.
1999 (mb = 5.8) at a distance of D = 13°. Note the clearly recognizable polarity of the first P-
wave half-cycle! The record components have been rotated into the directions Z, R and T
after determination of the backazimuth from first-motion polarities in Z, N and E (BAZ =
134°). Accordingly, P and Rayleigh waves are strongest in Z and R while S and Love wave
are strongest in T. Right: The recordings after SP bandpass filtering (0.5-5.0 Hz). The SNR
for the P-wave first-motion amplitude is much smaller and their polarity less clear. Also later
arrivals required for distance determination are no longer recognizable (signal processing
done with SEISAN; courtesy of L. Ottemöller).
Rough estimates of D may be made - in the absence of travel-time tables or curves or related
computer programs – using the following “rules-of- thumb”:
hypocenter distance d [in km] ≈ ∆t(Sg-Pg) [in s] × 8 (near range only) (11.1)
epicentral distance D [in km] ≈ ∆t(Sn-Pn) [in s] × 10 (in Pn-Sn range < 15°) (11.2)
epicentral distance D [in °] ≈ {∆t(S-P) [in min] - 2} × 10 (for 20°< D <100°) (11.3)
In the absence of travel-time curves or tables for depth phases one may use another “rule-of-
thumb” for a rough estimate of source depth from the travel-time difference ∆t(pP-P):
source depth h [in km] ≈ = ∆t(pP-P)/2 [in s] × 7 (for h < 100 km) (11.4)
or × 8 (for 100 km < h < 300 km)
or × 9 (for h > 300 km)
Bormann (1971a) showed that in the absence of a sufficiently strong P-wave arrival, the
backazimuth can be determined from horizontal components of any later seismic phase which
is polarized in the vertical propagation plane, such as PP, PS, PKP or SKS. These phases are
often much stronger in BB or LP records than P. However, because of phase shifts on internal
caustics (PP, PS, SP, PKPab) for most of these phases the 180° ambiguity in azimuth
determined from the ratio AE/AN can not be resolved as it can for P by taking into account the
first-motion polarity in the Z component. However, by considering the inhomogeneous global
distribution of earthquake belts, this problem can usually be solved.
Modern computer programs for seismogram analysis include subroutines that allow quick
determination of both azimuth and incidence angle from particle motion analysis over the
whole waveform of P or other appropriate phases. This is done by determining the direction
of the principal components of the particle motion, using, as a measure of reliability of the
calculated azimuth and incidence angle, the degree of particle motion linearity/ellipticity.
Such algorithms are available in the SEIS89 software (Baumbach, 1999). Christoffersson et
al. (1988) describe a maximum-likelihood estimator for analyzing the covariance matrix for a
single three-component seismogram (see also Roberts and Christoffersson, 1990). The
procedure allows joint estimation of the azimuth of approach, and for P and SV waves the
apparent angle of incidence and, hence, information on apparent surface velocity and thus on
epicentral distance. This was been implemented in the SEISAN software (Havskov and
Ottemöller, 1999). Fig. 11.25 shows an example of the application of the software to a
portion of the BB recording at Kongsberg (KONO) in Norway for the 12 November 1999,
Turkey earthquake (Mw = 7.1). The program finds a high correlation (0.9) between the
particle motions in the three components, gives the estimate of the backazimuth as 134°, an
apparent velocity of 9.6 km/s and the corresponding location of this earthquake at 40.54°N
and 30.86° E. This was only about 50 km off the true epicenter.
Applying similar algorithms to digital 3-component data from short-period P waves recorded
at regional distances, Walck and Chael (1991) show that more than 75% of the records
yielded backazimuth within 20° of the correct values. They found, however, a strong
dependence on the geological structure. Whereas stations located on Precambrian terranes
produced accurate backazimuth for SNR > 5 dB, stations on sedimentary rocks with
complicated structure had much larger errors. Excluding these stations, the RMS backazimuth
error is only about 6° for recordings with SNR > 10 dB.
Ruud et al. (1988) found that three-component locations for epicenters at distances up to
about 1000 km seldom deviated more than 50 km from network solutions, such deviations
being mainly due to errors in azimuth estimates. For short-period teleseismic P waves,
however, location errors occasionally exceeded 800 km, mainly because of poor distance
estimates derived from incidence angles (slowness) alone. For stronger sources, where BB
records can be used, distance can be determined using travel-time differences. The location
errors are then reduced to about 1°. Thus, three-component digital broadband data allow
reliable epicenters to be determined quickly with just single station records, and even data
from stations that still use analog recording may provide rapid and reliable epicenter
estimates. For combined single station and network location see Cassidy et al. (1990).
Fig. 11.25 Example of azimuth determination and epicenter location of the 12 Nov. 1999
Turkey earthquake by correlation analysis of three-component digital BB records at station
KONO, Norway. Backazimuth, apparent velocity, and correlation factor are determined from
the P-wave record section marked in the upper figure. For more details see text (signal
processing done with SEISAN; courtesy of L. Ottemöller).
11.2.7 Magnitude determination

When epicentral distance and depth of a seismic source are (at least roughly) known the
magnitude of the event can be estimated. The general procedures to be followed in magnitude
determination (and the measurement of amplitudes, periods or record duration) as well as the
specifics of different magnitude scales to be used for local, regional or teleseismic recordings
are dealt with in detail in section 3.2. DS 3.1 gives the magnitude calibration functions, both
for the teleseismic standard magnitudes (mb and Ms) and several other magnitude scales for
local, regional and teleseismic magnitudes. The various procedures can be learnt from an
exercise given in EX 3.1, which also gives solutions for the different tasks.
11.3 Routine signal processing of digital seismograms
11.2.8 Hypocenter location by means of network and array recordings
Hypocenter location is simplified if records from at least 3 stations are available. The more
uniformly the stations are distributed around the source in azimuth and distance (with
distances ranging from close-in to long range) and the more seismic phases are used for
location, the lower the uncertainty in the estimates. The procedures in both manual and
computer assisted multi-station hypocenter location are outlined in IS 11.1, which gives the
underlying algorithms and error calculations, as well as standard and advanced methods for
both absolute and relative location. Also discussed is the influence of deviations from the
assumed Earth models on the locations. The improvements in hypocenter relocation
achievable with better Earth models are also demonstrated.
EX 11.1 aims at epicenter location by a simple circle and chord method using seismograms
from local stations both inside and outside the network. The epicentral distances have to be
determined first for each station by identifying on its records the phases Pg, Sg, Pn and/or Sn
and matching them to a local travel-time curve.
With digital multi-station data and advanced seismogram analysis software, source location
becomes almost a trivial task. One just picks a sufficient number of first arrival times (see Fig.
11.5), activates the relevant location program for local, regional and/or teleseismic sources
and gets the result, including a map showing the epicenter if required, in an instant. The
accuracy of location, particularly source depth, can be significantly improved by picking not
only P-wave first arrivals but later arrivals too, which give a much better distance and depth
control than slowness data alone. Examples for both local and teleseismic event locations
based on seismic network and array data are given in the following sections. Location using
array data is described in Chapter 9, together with the underlying theory.

Standard analysis includes all data pre-processing and processing operations for the
interpretation and inversion of broadband seismograms. Important time-domain processes are
signal detection, signal filtering, restitution and simulation, phase picking, polarization
analysis as well as beamforming and vespagram analysis for arrays. In the frequency domain
the main procedures are frequency-wavenumber (f-k) and spectral analysis. Array-techniques
as f-k and vespagram analysis, slowness and azimuth determination for plane waves, and
beamforming are discussed in detail in Chapter 9 but a few examples are also shown below.
Spectral analysis can be used for the estimation of the frequency content of a seismic wave,
and of seismic noise (see 4.1 and 7.2, respectively).
11.3.1 Signal detection

The first task of routine data analysis is the detection of a seismic signal. A signal is
distinguishable from the seismic background noise either on the basis of its larger amplitudes
or its differences in shape and frequency. Various methods are used for signal detection.
Threshold detectors and frequency-wavenumber analysis are applied to the continuous stream
of data. In practice, the threshold is not constant but varies with the season and the time of
the day. For this reason, the threshold detectors determine the average signal power in two
moving time windows: one long term (LTA) and one short term (STA). The ratio of the STA
to LTA corresponds to the signal-to-noise-ratio (SNR). For details on the STA/LTA trigger
and its optimal parameter setting see IS 8.1.
In practice, BB records are filtered before detectors are used. Useful filters are Butterworth
high-pass filters with corner frequencies fc > 0.5 Hz or standard band-pass types with center
frequency f = 1 Hz for teleseismic P waves and high-pass filter with fc > 1 Hz for local
sources. Fig. 11.26 demonstrates detection and onset-time measurement for a weak, short-
period P wave. In the lowermost 30 s segment of a BB-velocity seismogram the oceanic
microseisms dominate in the period band 4-7 s. The two other traces are short-period
seismograms after narrow band-pass (BP) filtering with: (1) a filter to simulate a WWSSN-SP
seismogram; and (2) a two step Butterworth BP filter of 2nd-order with cut-off frequencies of
0.7 and 2 Hz, respectively. The latter filter produces, for the noise conditions at the GRF-
array, the best SNR for teleseismic signals. Seismic networks designed to detect mainly local
seismic events may require other filter parameters that take account of local noise conditions,
for optimal detection (see IS 8.1).
Generally, a seismic signal is declared when the SNR exceeds a pre-set threshold. Various
procedures, some analytical and some based on personal experience, are used to differentiate
between natural earthquakes, mining-induced earthquakes and different kinds of explosions.
Usually, the detected signals are analyzed for routine parameter extraction and data exchange.
Fig 11.26 Bandwidth and SNR: A small short-period P-wave arrival which is within the noise level on a BB-
velocity record (lower trace) may be detected by using a WWSSN-SP simulation filter (middle trace) or a
Butterworth band-pass filter (BP; uppermost trace). The SNR is 0.2 on the original BB record, about 1 on the
WWSSN-SP filter and about 2 on the BP-filtered trace. The seismogram is of an earthquake in the Kurile Islands
on 25 March 2002, 6:18:13 UT, recorded at station GRA1, Germany.
11.3.2 Signal filtering, restitution and simulation

Classical broadband seismographs, such as the Russian Kirnos SKD, record ground
displacement with constant magnification over a bandwidth of 2.5 decades or about 8 octaves.
The IDA-system (International Deployment of Accelerometers) deployed in the 1970s, used
originally LaCoste-Romberg gravimeters for recording long-period waves from strong
earthquakes proportional to ground acceleration over the band from DC to about 0.1 Hz
(nowadays replaced by STS1). Modern strong-motion sensors such as the Kinemetrics Inc.
Episensor ES-T have a flat response to ground acceleration in an even broader frequency band
from DC to 200 Hz. In contrast, feedback-controlled BB sensors for recording weak-motion
usually have a response proportional to the ground velocity (see Fig. 11.27 right). Such BB
recordings, however, are often not suitable for direct visual record analysis and parameter
extraction in the time domain. Low-frequency signals and surface waves of weak earthquakes
are not or only poorly seen. Therefore, BB data must be transformed by applying digital
filters in a way that yields optimal seismograms for specific investigations and analysis.
For some research tasks and ordinary routine analysis of BB seismograms the application of
high-pass, low-pass and band-pass filters is usually sufficient. However, simultaneous multi-
channel data processing or the determination of source parameters according to internationally
agreed standards (such as body- and surface-wave magnitudes, which are defined on the basis
of former analog band-limited recordings) often require simulation of a specific response,
including those of classical analog seismograph systems (Seidl, 1980). Another special
problem of simulation is “restitution”. Restitution is the realization of a seismograph system
whose transfer function is directly proportional to ground displacement, velocity or
acceleration in the broadest possible frequency range. The restitution of the true ground
displacement down to (near) zero frequencies is a precondition for seismic moment-tensor
determinations both in the spectral and the time domain (e.g., signal moment; see Fig. 11.6).
It is achieved by extending the lowermost corner frequency of the seismometer
computationally far beyond that of the physical sensor system. Both the simulation of
arbitrary band-limited seismograph systems as well as the extreme broadband “restitution” of
the true ground motion is therefore a necessary step in pre-processing of digital BB data.
Simulation is the mapping of a given seismogram into the seismogram of another type of
seismograph, e.g., those of classical analog recordings such as WWSSN-SP, WWSSN-LP,
Kirnos SKD, SRO-LP, and Wood-Anderson (WA). Up to now, amplitudes and periods for the
determination of body- and surface-wave magnitudes mb and Ms are measured on simulated
WWSSN-SP and WWSSN-LP or SRO-LP seismograms, respectively, and the maximum
amplitude for the original local Richter magnitude is measured on Wood-Anderson simulated
seismograms. Fig. 11.27 (left) depicts the displacement response of these seismographs.
The possibility of carrying out these simulations with high accuracy and stability defines the
characteristics that have to be met by modern digital broadband seismograph:
• large bandwidth;
• large dynamic range;
• high resolution;
• low instrumental seismometer self-noise (see 5.6.2);
• low noise induced by variations in air pressure and
temperature (see 5.3.4, 5.3.5, and 7.4.4);
• analytically exactly known transfer function (see 5.2).
Fig. 11.27 (right) depicts the displacement responses of a few common BB-velocity sensors
such as:
• the original Wielandt-Streckeisen STS1 with a bandwidth of 2 decades between the
3-db roll-off points at frequencies of 0.05 Hz and 5 Hz (anti-aliasing filter). These
seismographs are deployed in the world’s first broadband array (GRF) around
Gräfenberg/Erlangen in Germany (see Fig. 11.3a);
• the advanced STS1 that is generally used at the global IRIS network of very
broadband (VBB) stations (velocity bandwidth of about 3.3 decades between 5 Hz
and 360 s; see also DS 5.1);
• the STS2 seismographs (see DS 5.1) that are usually operated in the frequency range
between 0.00827 Hz and 40 Hz (velocity bandwidth of 3.7 decades or about 12
octaves, respectively). They are used at the stations of the GRSN (see Fig. 11.3a) but
also deployed world-wide at stations of the GEOFON network and at many others.
All these seismographs can be considered to be linear systems within the range of their usual
operation. The transfer function H(s) of a linear system can be calculated from its poles and
zeros by using the following general equation:
H(s) = N * Π (s – zi) / Π (s – pk) (11.5)
where N is the gain factor, s = jω with ω = 2πf and j the complex number √-1, zi are the zeros
numbering from i = 1 to m and pk the poles with k = 1 to n. Zeros are those values for which
the numerator in Eq. (11.5) becomes zero while the poles are the values for which the
denominator becomes zero.
Tab. 11.3 summarizes the poles and zeros of the classical standard responses WWSSN-SP,
WWSSN-LP, WA (Wood-Anderson), Kirnos SKD and SRO-LP which control the shape of
the response curves. Tab. 11.4 gives the same for the three broadband responses shown in Fig.
11.27 on the right. Not given are the gain factors because they depend on the specific data
acquisition system and its sensitivity.
Fig. 11.27 Left: Displacement amplitude response characteristics of classical seismographs;

right: The same for broadband seismographs STS1(GRF) (old version as used at the
Gräfenberg array), STS1 (VBB) (advanced version as used in the IRIS global network) and
STS2. For STS1 (VBB) and STS2 no anti-aliasing filter is shown. The classical responses
shown on the left can be simulated with digital data from these broadband systems (see text).
Tab. 11.3 Zeros and poles corresponding to the displacement transfer functions depicted in
Fig. 11.27 left for the classical analog standard seismographs WWSSN-SP, WWSSN-LP,
WA, Kirnos SKD and SRO-LP.
Seismograph Zeros Poles

WWSSN-SP (0.0, 0.0) (-3.3678, -3.7315) (=p1)
(0.0, 0.0) (-3.3678, 3.7315) (=p2)
(0.0, 0.0) (-7.0372, -4.5456) (=p3)
(-7.0372, 4.5456) (=p4)
WWSSN-LP (0.0, 0.0) (-0.4189, 0.0)
(0.0, 0.0) (-0.4189, 0.0)
(0.0, 0.0) (-6.2832E-02, 0.0)
(-6.2832E-02, 0.0)
WA (0.0, 0.0) (-6.2832, -4.7124)
(0.0, 0.0) (-6.2832, 4.7124)
Kirnos SKD (0.0, 0.0) (-0.1257, -0.2177)
(0.0, 0.0) (-0.1257, 0.2177)
(0.0 0.0) (-80.1093, 0.0)
(-0.31540, 0.0)
SRO-LP (0.0, 0.0) (-1.3000E-01, 0.0)
(0.0, 0.0) (-6.0200, 0.0)
(0.0, 0.0) (-8.6588, 0.0)
(-5.0100E+01, 0.0) (-3.5200E+01, 0.0)
(-0.0, 1.0500) (-2.8200E-01, 0.0)
(-0.0, -1.0500) (-3.9300, 0.0)
(0.0, 0.0) (-2.0101E-01, 2.3999E-01)
(0.0, 0.0) (-2.0101E-01, -2.3999E-01)
(-1.3400E-01, 1.0022E-01)
(-1.3400E-01, -1.0022E-01)
(-2.5100E-02, 0.0)
(-9.4200E-03, 0.0)
Tab. 11.4 Zeros and poles corresponding to the displacement transfer functions of the
velocity-proportional broadband seismographs STS1(GRF), STS1-VBB(IRIS) and STS2 as
depicted in Fig. 11.27 right. From their output data seismograms according to the classical
analog standard seismographs WWSSN-SP, WWSSN-LP, WA, Kirnos SKD and SRO-LP are
routinely simulated at the SZGRF in Erlangen, Germany.
Seismograph Zeros Poles

STS2 (0.0, 0.0) (-3.674E-2, -3.675E-3)
(0.0, 0.0) (-3.674E-2, 3.675E-3)
(0.0, 0.0)
STS1(GRF) (0.0, 0.0) (-0.2221, -0.2222)
(0.0, 0.0) (-0.2221, 0.2222)
(0.0, 0.0) (-31.416, 0.0)
(-19.572, 4.574)
(-19.572, -24.574)
(-7.006, 30.625)
(-7.006, -30.625)
(-28.306, 13.629)
(-28.306, -13.629)
STS1(VBB)) (0.0, 0.0) (-1.2341E-02, 1.2341E-02)
(0.0, 0.0) (-1.2341E-02, -1.2341E-02)
(0.0, 0.0)
Using the data given in these tables, the exact responses of the respective seismographs can be
easily found. As an example, we calculate the response curve of the WWSSN-SP. According
to Tab. 11.3 it has three zeros and four poles. Thus we can write Eq. (11.5) as
H(s) = N * s3 / (s-p1)(s-p2)(s-p3)(s-p4) (11.6)

with
p1 = -3.3678 – 3.7315j
p2 = -3.3678 + 3.7315j
p3 = -7.0372 – 4.5456j
p4 = -7.0372 + 4.5456j.
Taking into account the discussions in section 5.2.7, the squared lower angular corner
frequency of the response (that is in the given case the eigenfrequency of the WWSSN-SP
seismometer) is ωl2 = p1⋅p2 whereas the squared upper angular eigenfrequency (which used to
be in the classical SP records that of the galvanometer) is ωu2 = p3⋅p4. Since the product of
conjugate complex numbers (a + bj) (a – bj) = a2 + b2 it follows for the poles:
ωl2 = 25.27 with fl = 0.80 Hz and

ωu2 = 70.18 with fu = 1.33 Hz.
When comparing these values for the corner frequencies of the displacement response of
WWSSN-SP in Fig. 11.27 (left) one recognizes that the maximum displacement
magnification (slope approximately zero) lies indeed between these two values. Further, as
outlined in 5.2.7, a conjugate pair of poles such as p1 and p2 or p3 and p4 correspond to a
second order corner of the amplitude response, i.e., to a change in the slope of the asymptote
to the response curve by 2 orders. Further, the number of zeros controls the slope of the
response curve at the low-frequency end, which is three in the case of the WWSSN-SP (see
Eq. (11.6) and Tab. 11.3). Thus, at its low-frequency end, the WWSSN-SP response has
according to its three zeros a slope of 3. This changes at the first pair of poles, i.e., at fl = 0.8
Hz, by 2 orders from 3 to 1 (i.e., to velocity proportional!), and again at fu = 1.33 Hz by two
orders from 1 to –1. This is clearly to be seen in Fig. 11.27. In the same manner, the general
shape of all the responses given in that figure can be assessed or precisely calculated
according to Eq. (11.5) by using the values for the poles and zeros given in Tabs. 11.3 and
11.4. Doing the same with the values given in Tab. 11.3 for WWSSN-LP one gets for fl =
0.06667 Hz, corresponding to the 15 s seismometer and fu = 0.009998 Hz corresponding to
the 100 s galvanometer, used in original WWSS-LP seismographs. The aim of the exercise in
EX 5.5 is to calculate and construct with the method shown above the responses of
seismographs operating at several seismic stations of the global network from the data given
in their SEED header information.
Note that the poles and zeros given in Tabs. 11.3 and 11.4 are valid only if the input signal to
the considered seismographs is ground displacement (amplitude Ad). Consequently, the values
in Tab. 11.3 are not suitable for simulating the responses of the classical seismographs if the
input signal to the filter is not displacement. From the output of the STS2, any simulation
filter gets as an input a signal, which is velocity-proportional within the frequency range
between 0.00827 Hz and 40 Hz. Its amplitude is Av = ωAd. Accordingly, the transfer function
of the simulation filter Hfs(s) has to be the convolution product of the inverse of the transfer
function Hr(s) of the recording instrument and the transfer function Hs(s) of the seismograph
that is to be simulated:
Hfs (s) = Hr -1(s) * Hs(s). (11.7)
Thus, even for the same Hs(s) to be simulated, the poles and zeros of the simulation filter
differ depending on those of the recording seismograph. Tab. 11.5 gives, as an example from
the SZGRF, the poles and zeros of the displacement filters for simulating the responses shown
in Fig. 11.27 (left), and the poles and zeros given in Tab. 11.3, from output data of the STS2.
Tab. 11.5 Poles and zeros of the simulation filters required for simulating standard
seismograms of WWSSN-SP, WWSSN-LP, WA, Kirnos SKD and SRO-LP, respectively
from STS2 BB-velocity records.
Simulation- Zeros Poles

filter for
WWSSN-SP (-3.6743E-02, -3.6754E-02) (-3.3678, -3.7316)
(-3.6743E-02, 3.6754E-02) (-3.3678, 3.7315)
(-7.0372, -4.5456)
(-7.0372, 4.5456)
WWSSN-LP (-3.6743E-02, -3.6754E-02) (-0.4189, 0.0)
(-3.6743E-02, 3.6754E-02) (-0.4189, 0.0)
(-6.2832E-02, 0.0)
(-6.2832E-02, 0.0)
WA (-3.6743E-02, -3.6754E-02) (-6.2832, -4.7124)
(-3.6743E-02, 3.6754E-02) (-6.2832, 4.7124)
(0.0, 0.0)
Kirnos SKD (-3.6743E-02, -3.6754E-02) (-0.12566, -0.2177)
(-3.6743E-02, 3.6754E-02) (-0.1257, 0.2177)
(-80.1094, 0.0)
(-0.3154, 0.0)
SRO-LP (-3.6744E-02, -3.6754E-02) (-1.3000E-01, 0.0)
(-3.6743E-02, 3.6754e-02) (-6.0200, 0.0
(-5.0100E+01, 0) (-8.6588, 0.0)
(-0, 1.0500) (-3.5200E+01, 0.0)
(-0, -1.0500) (-2.8200E-01, 0.0)
(0.0, 0.0) (-3.9301E+00, 0.0)
(0.0, 0.0) (-2.0101E-01, 2.3999E-01)
(-2.0101E-01, -2.3999E-01)
(-1.3400E-01, 1.0022E-01)
(-1.3400E-01, -1.0022E-01)
(-2.5100E-02,0.0)
(-9.4200E-03,0.0)
Fig. 11.28 shows a comparison of the original three-component BB-velocity record of an

STS2 at station WET from a local earthquake in Germany with the respective seismograms of
a simulated Wood-Anderson (WA) seismograph. For a teleseismic earthquake Fig. 11.29
gives the STS2 BB-velocity record together with the respective simulated records for
WWSSN-SP and LP. Figs. 11.30 and 11.31 give two more examples of both record
simulation and the restitution of very broadband (VBB) true ground displacement. VBB
restitution of ground displacement is achieved by convolving the given displacement
response of the recording seismometer with its own inverse, i.e.,:
Hrest(s) = Hs -1(s) * Hs(s). (11.8)

However, Eq. (11.8) works well only for frequencies smaller than the upper corner frequency
(anti-alias filter!) and for signal amplitudes that are well above the level of ambient, internal
(instrumental), and digitization noise.
Fig. 11.28 3-component recordings at station WET (Wettzell) of a local earthquake at an

epicentral distance of D = 116 km. Lower traces: original STS2 records with sampling rate of
80 Hz; upper traces: simulated Wood-Anderson (WA) recordings. Note that the displacement-
proportional WA record contains less high frequency oscillations than the velocity-
proportional STS2 record (compare responses shown in Fig. 11.27).
Fig. 11.29 BB-velocity seismogram (top) and simulated WWSSN-SP (middle) and WWSSN-
LP seismograms (bottom). Note the strong dependence of waveforms and seismogram shape
on the bandwidth of the simulated seismographs.
Both Fig.11.29 and 11.30 show clearly the strong influence of differences in bandwidth and
center frequencies of the seismometer responses (compare with Fig. 11.27) on both the
individual waveforms and the general shape of the seismogram. This is particularly obvious in
the simulated teleseismic earthquake records. Fig. 11.30 shows the recordings of the
teleseismic P-wave group of an earthquake in California on 16 Sept. 1999. Shown are the
restitution of a BB-displacement seismogram derived from a BB-velocity seismogram and the
simulations of WWSSN-SP and SRO-LP seismograms. In the BB-velocity seismogram one
recognizes clearly the superposition of a low-frequency signal and a high-frequency wave
group. The latter is clearly seen in the WWSSN-SP record but is completely absent in the
SRO-LP simulation. From this comparison it is obvious that both the BB-velocity and the SP
seismograms enhance short-period signal amplitudes. Therefore, only the former recordings
are well suited for studying the fine structure of the Earth and determining the onset time and
amplitude of short-period P waves. In contrast, BB-displacement seismograms and LP-
filtered seismograms suppress the high-frequencies in the signals. Generally they are more
suited to routine practice for surface-wave magnitude estimation and for the identification of
most (but not all!) later phases (see Figs. 11.12, 11.13, 11.37, and Fig. 2.23).
Fig. 11.30 From top to bottom: The original BB-velocity seismogram recorded at station
GRFO; the BB-displacement record derived by restitution; the simulated WWSSN-SP; and
the simulated SRO-LP seismograms of the P-wave group from an earthquake in California
(16 Sept. 1999; D = 84.1°; Ms = 7.4).
Fig. 11.31 shows 10-days of a VBB record from an STS1 vertical-component seismograph
(corner period Tc = 360s) at station MOX and simulated WWSSN-SP and SRO-LP
seismograms for a short (40 min) time segment of this VBB record.
Fig. 11.31 STS1 (Tc = 360s) vertical-component seismogram with a length of 10 days (upper
trace) as recorded at MOX station, Germany. In the seismogram we recognize Earth's tides
and different earthquakes as spikes. For one of these earthquakes a WWSSN-SP and SRO-LP
simulation filter was applied (lower traces). The length of the filtered records is 40 minutes.
Figs. 11.32a-d demonstrate, with examples from the GRSN and the GRF array in Germany,
the restitution of (“true”) displacement signals from BB-velocity records as well as the
simulation of WWSSN-SP, Kirnos BB-displacement and SRO-LP records. All traces are
time-shifted for the P-wave group and summed (they are aligned on trace 16). The summation
trace forms a reference seismogram for the determination of signal form variations. Generally,
this trace is used for the beam (see 11.3.5 below). The different records clearly demonstrate
the frequency dependence of the spatial coherence of the signal. Whereas high-frequency
signals are incoherent over the dimension of this regional network (aperture about 500 to 800
km) this is not so for the long-period records which are nearly identical at all recording sites.
The following features are shown in Figs. 11.32a-d:

a) Time shifted BB-displacement (traces 16-30) and BB-velocity seismograms (traces 1-15)
with a duration of 145 s of the P-wave group from an earthquake in Peru on 23 June 2001
(Ms=8.1) as recorded at 15 stations of the GRSN. The BB-displacement seismogram
suppresses the high-frequencies, which are clearly shown on the BB-velocity record.
b) WWSSN-SP simulations for the same stations as in Figure 11.32a. The high-frequency
signals are enhanced but the shape and amplitudes of the waveforms are shown to vary
considerably within the network, i.e., the coherence is low.
c) Kirnos SKD BB-displacement and d) SRO-LP simulations for the same stations as in
Fig.11.32a. The high-frequency signals are masked. All traces show coherent waveforms.
Fig. 11.32a (for explanation see text on page 48)
Fig. 11.32b As Fig. 11.32a however with short-period WWSSN-SP simulation.

Fig 11.32c As Fig. 11.32a but for displacement-proportional Kirnos SKD simulation.
Fig. 11.32d As Fig. 11.32a but for long-period SRO-LP simulation.

Fig. 11.32a-d Restitution, simulation and coherency of seismograms demonstrated with
records of the GRSN from an earthquake in Peru (23 June 2001, Ms = 8.1) in the epicentral
distance range from 96° to 100°; for explanation see text).
11.3.3 Signal coherency at networks and arrays

Heterogeneous crustal structure and the array aperture limit the period band of spatially
coherent signals. The larger the aperture of an array the more rapidly the signal coherence
falls off with frequency. At short periods the array behaves like a network of single stations
whereas at long periods the array behaves like a sensitive single station. For the GRF-array
(aperture about 50 to 120 km) for instance, the signals are coherent for periods between about
1 and 50 s. For the GRSN the band of coherent signals is at longer periods than for the small
aperture detection arrays like GERES in Germany or NORES in Norway (aperture 4 and 3
km, respectively) where signals are coherent at periods shorter than 1 s. In the coherency band
itself, waveforms vary depending on their dominant frequency, apparent horizontal velocity
and azimuth of approach. For instance, coherent waveforms are observed from the GRSN for
BB-displacement records, Kirnos SKD simulation and all long-period simulated seismograms
(see Figs. 11.32a-d) whereas for simulated WWSSN-SP seismograms the waveforms have
low coherence or are incoherent.
Figs. 11.33a and b shows a comparison of the first 14 s of the P wave of the GRSN and the
GRF-array. The coherence is clearly higher in the short-period range for the recordings at the
smaller GRF-array than for the GRSN. The GRSN works as an array for periods longer than
about 10 s but it is a network for shorter periods where the GRF-array works as an array down
to periods of about 1 s. This discussion is valid for teleseismic signals only, where the
epicentral distance is larger than the aperture of the station network or array.
Fig. 11.33a (see figure caption below)

Fig. 11.33b (see figure caption below)
Fig. 11.33 WWSSN-SP simulations of the first 14 s after the P-wave onset from the same
Peru earthquake as in Fig. 11.32. a) Recordings at the GRSN. b) Recordings at the GRF
array. Note the lower coherence of the waveforms recorded at the stations of the regional
network, which has an aperture much larger than the GRF-array (see Fig. 11.3a). The
summation traces 17 and 14, respectively, are reference seismograms for the determination of
signal waveform variations.
11.3.4 f-k and vespagram analysis

Array-techniques such as f-k and vespagram analysis should be applied only to records with
coherent waveforms. Vespagram analysis or the velocity spectrum analysis is a method for
separating signals propagating with different apparent horizontal velocities. The seismic
energy reaching an array from a defined backazimuth with different slownesses is plotted
along the time axis. This allows identification of later phases based on their specific slowness
values. The best fitting slowness is that for which a considered phase has the largest
amplitude in the vespagram. Fig. 11.34 shows the original records from the GRSN (top) and
the related vespagram (bottom). More vespagrams are given in Figures 12e–g of DS 11.2.
Fig. 11.34 Top: Simulated vertical-component WWSSN-LP seismograms from an

earthquake in the region of Papua New Guinea. Source data NEIC-QED: 10 May 1999; depth
137 km; mb = 6.5; D = 124° to GRF, BAZ = 51°. The phases Pdif (old Pdiff), PKPdf,
pPKPdf, PP, pPP, sPP and an unidentified phase X have been marked. Bottom: Vespagram of
the upper record section. The analysis yields slowness values of 4.5s/° for Pdif, 2.0s/° for
PKPdf and pPKPdf, 7.0s/° for PP, and a value that corresponds to the slowness for Pdif for
the unknown phase X.
As an example an f-k analysis is shown in Fig. 11.35. The f-k analysis is used to determine
slowness and backazimuth of coherent teleseismic wave groups recorded at an array. The
epicentral distance must be much larger than the aperture of the recording array. The f-k
analysis transforms the combined traces within a current time window (Fig. 11.35a and b) into
the frequency-wavenumber domain. The result in the f-k domain is displayed in a separate
window (Fig. 11.35c) with amplitudes (corresponding to wave energy) coded in color. A good
result is achieved when there is a single, prominent color in the maximum (yellow in Fig.
11.35c). This maximum denotes slowness and backazimuth of the investigated phase and is
helpful for source parameter determination and phase identification. The example was
recorded at the GRF-array from an earthquake in Novaya Zemlya. Slowness and backazimuth
values are 7.3 s/° and 11°, respectively. These values are used for producing the beam.
Fig. 11.35a
Figs. 11.35a-c Illustration of the procedure of frequency-wavenumber (f-k) analysis: a)
coherent P-wave signals recorded at the GRF-array stations from an earthquake on 19 April
1997 (the box marks the time window selected for the f-k-analysis); b) the zoomed window
used for the f-k-analysis; c) energy (coded in colors) in the frequency range 0.39-2.97 Hz as a
function of wavenumber k. A good result is achieved because the single, prominent maximum
(in yellow) shows the presence of a coherent signal. The estimated slowness and backazimuth
values are 7.3 s/° and 11°, respectively.
Fig. 11.35b
Fig. 11.35c
11.3.5 Beamforming
Beamforming improves the SNR of a seismic signal by summing the coherent signals from
array stations (see 9.4.5). Signals at each station are time shifted by the delay time relative to
some reference point or station. The delay time depends on ray slowness and azimuth and can
be determined by trial and error or by f-k analysis. The delayed signals are summed “in
phase” to produce the beam. Fig. 11.35d presents the array recordings of the signal shown in
Figs. 11.35a-b time shifted, to correct for the time delay, and summed (trace 14; beam). In
delay-and-sum beamforming with N stations the SNR improves by a factor √N if the noise is
uncorrelated between the seismometers. In the summation the increase in amplitude of the
coherent signal is proportional to N. For incoherent waves (random seismic noise in
particular), it is only proportional to √N. Thus, f-k analysis and beamforming are helpful for
routine analysis if very weak signals have to be detected and analyzed.
Fig. 11.35d The delay times for each of the array seismometers have been calculated from
the slowness and azimuth of the signals shown in Figs. 11.35a-c. The time-shifted signals are
summed “in phase” to produce the beam (top trace) which has the better signal-to-noise ratio
than the recordings from the individual seismometers.
Fig. 11.36 shows another example of array processing with short-period filtered seismograms
of the GRF array. The signal on the beam trace is the PKP wave of an underground nuclear
explosion at Mururoa Atoll with an explosion yield equivalent of about 1 kt TNT. The onset
time and signal amplitude of a weak seismic signal can only be read on the beam. The peak-
to-peak amplitude is only about 2 nm with a period around 1 s.
Fig. 11.36 Detection of the PKP wave of a nuclear explosion at Mururoa Atoll on 27 June
1982 at the Gräfenberg array using the delay-and-sum-method and a very narrowband
Butterworth bandpass filter (BP) centered around 1 Hz. The event occurred at an epicentral
distance of 146° and the explosion yield was approximately 1 kt TNT.
11.3.6 Polarization analysis

The task of polarization analysis is the transformation of recorded three component
seismograms into the ray-oriented co-ordinate systems. For linearly polarized and single pulse
P waves in a lateral homogeneous Earth, this task is simple, at least for signals with a high
SNR; the direction of the polarization vector of the P wave clearly determines the orientation
of the wave co-ordinate system. However, when propagating through heterogeneous and
anisotropic media, the seismic waves have three-dimensional and frequency-dependent
particle motions and the measured ray-directions scatter by ten degrees and more about the
great circle path from the epicenter to the station (see Fig. 2.6).
Determination of particle motion is included in most of the analysis software. For

identification of wave polarization and investigation of shear-wave splitting, the rotation of
the traditional components N, E, and Z into either a ray–oriented co-ordinate systems or into
the directions R (radial, i.e., towards the epicenter) and T (transverse, i.e., perpendicular to
the epicenter direction) is particularly suitable for the identification of secondary later phases.
An example for the comprehensive interpretation of such phases in a teleseismic record is
given in Fig. 11.37.
Fig. 11.37 Simulation of three-component WWSSN-LP seismograms of the Volcano Islands
earthquake of 28 March 2000, recorded at station RUE in Germany (D = 94°, h = 119 km).
The horizontal components N and E are rotated into R and T components. The phases P, pP,
SP and the beginning of the dispersed surface Rayleigh wave train LR are marked on the
vertical-component seismogram, SKS, PS on the radial component (R) and S, SS, SSS and
the beginning of the Love waves LQ on the transverse component (T), respectively. Not
marked (but clearly recognizable) are the depth phases sS behind S, sSS behind SS, and
SSSS+sSSSS before LQ. The record length is 41 min.
11.4 Software for routine analysis

11.4.1 SHM
The Seismic Handler SHM is a powerful program for analyzing local, regional and
teleseismic recordings. K. Stammler of the SZGRF in Erlangen has developed it for the
analysis of data from the Graefenberg (GRF) array and the German Regional Seismic
Network (GRSN). The program and descriptions are available via http://www.szgrf.bgr.de/sh-
doc/index.html.
Main features of the program are:

• application of array procedures to a set of stations (slowness- and backazimuth
determination by means of beamforming and f-k analysis);
• location algorithms (teleseismic locations using travel-time tables and empirical
correction vectors, local and regional locations via external programs, e.g.,
LocSAT).
11.4 Software for routine analysis
The basic program has some (more or less) standardized options, e.g.:
• manual and automatic phase picking (see Fig. 11.5);
• trace filtering with simulation and bandpass filters (see Figs. 11.28 and 11.29);
• determination of amplitudes, periods and magnitudes (see Fig. 11.4);
• display of theoretical travel times on the traces (see Fig. 11.37).
Furthermore, the following tasks are implemented:

• rotation of horizontal components (see Figs. 11.13 and 11.37);
• particle motion diagrams (see Fig. 2.6);
• trace amplitude spectrum (see Fig. 11.47);
• vespagram trace display (see Fig. 11.34);
• determination of signal/noise ratio (see Fig. 11.47); and
• trace editing functions.
Different data formats are supported on continuous data streams of single stations, networks
and/or array stations. SHM is currently supported on UNIX and Linux. A screen display of
SHM is shown in Fig. 11.38.
Fig. 11.38 Screen display of the seismic analysis program SHM. Different windows display a
number of station recordings (large window), a zoomed single-station window, two
seismogram and source parameter windows (left side) and an output window for the results of
the seismogram analysis. Generally, the resulting parameters are stored in a database.
11.4.2 SEISAN
Another widely used seismic analysis system is SEISAN developed by J. Havskov and L.
Ottemöller (1999). It contains a complete set of programs and a simple database for
analyzing analog and digital recordings. SEISAN can be used, amongst other things, for phase
picking, spectral analysis, azimuth determination, and plotting seismograms. SEISAN is
supported by DOS, Windows95, SunOS, Solaris and Linux and contains conversion programs
for the most common data formats. The program, together with a detailed Manual, is available
via http://www.ifjf.uib.no/seismo/software/seisan/seisan.html.
11.4.3 PITSA
F. Scherbaum, J. Johnson and A. Rietbrock wrote the current version. It is a program for
interactive analysis of seismological data and has numerous tools for digital signal processing
and routine analysis. PITSA is currently supported on SunOS, Solaris and Linux and uses the
X11 windowing system. It is available via http://lbutler.geo.uni-potsdam.de/service.htm.
11.4.4 GIANT
Andreas Rietbrock has written this program package. It is a system for consistent analysis of
large, heterogeneous seismological data sets. It provides a graphical user interface (GUI)
between a relational database and numerous analysis tools (such as HYPO71, FOCMEC,
PREPROC, SIMUL, PITSA, etc. ). The GIANT system is currently supported on SunOS,
Solaris and Linux and uses the X11 windowing system and available via
http://lbutler.geo.uni-potsdam.de/service.htm.
11.4.5 Other programs and ORFEUS software links

C. M. Valdés wrote the interactive analysis program PCEQ for IBM compatible PCs. It is
widely used in conjunction with the location program HYPO71 for local events. The principal
features are: picking P- and S-wave arrivals; filtering the seismogram for better P- and S-
wave picks, and computing the spectra of selected seismogram sections. It is published in
Volume 1 of the IASPEI Software Library (Lee, 1995).
Andrey Petrovich Akimov has written the program WSG (in English AWP: Automated
workplace of seismologists), version 4.5 (in Russian). It works in an environment of Windows
95/98/NT and is used at single stations and seismic networks for estimating parameters from
local, regional and teleseismic sources. The program converts different seismic data formats
such as XDATA, PCC-1, CSS 2.8 and 3.0, DASS, CM6 GSE2 and can import via the TCP/IP
protocol data from NRTS and LISS systems (miniSEED). The program and the program
documentation in Russian is available via http://www2.gsras.ru/engl/mainms.htm.
ORFEUS (http://orfeus.knmi.nl/) presents a comprehensive list and links to available software

in seismology. It concentrates on shareware. However, some relevant commercial sites are
also included. Emphasis is laid on programs which run on UNIX/Linux platform.
11.5 Examples of seismogram analysis

The character of a seismogram depends on the source mechanism, the source depth, and
whether the epicenter of the source is at local, regional or teleseismic distances. Seismograms
of local earthquakes are characterized by short duration of the record from a few seconds to
say one minute, higher frequencies, and a characteristic shape of the wave envelope, usually
an exponential decay of amplitudes after the amplitude maximum, termed “coda” (see Figure
1b in DS 11.1 and Figure 2 in EX 11.1). In contrast, records at teleseismic distances show
lower frequencies (because high-frequency energy has already been reduced by anelastic
attenuation and scattering), and have a duration from say fifteen minutes to several hours (see
Fig. 1.2). Regional events have intermediate features. The various wave groups, arriving at a
station over different path, are called phases. They have to be identified and their parameters
determined (onset time, amplitude, period, polarization, etc.). Phase symbols should be
assigned according to the IASPEI recommended standard nomenclature of seismic phases.
For phase names, their definition and ray paths see IS 2.1. Fig. 11.39 shows seismograms
recorded at local, regional and teleseismic distances. They illustrate how the characteristics of
seismic records vary with distance and depending on the source type. These characteristics
will be discussed in more detail in the following sections.
There is no unique standard definition yet for the distance ranges termed “near” (“local” and
“regional”), or “distant” (“teleseismic”; sometimes subdivided into “distant” and “very
distant”). Regional variations of crustal and upper-mantle structure make it impossible to
define a single distance at which propagation of local or regional phases stops and only
teleseismic phases will be observed. In the following we consider a source as local if the
direct crustal phases Pg and Sg arrive as first P- and S-wave onsets, respectively. In contrast,
the phases Pn and Sn, which have their turning point in the uppermost mantle, are the first
arriving P and S waves in the regional distance range. However, as discussed in 2.6.1 and
shown in Fig. 2.40, the distance at which Pn takes over as first arrival depends on the crustal
thickness, average wave speed and the dip of the crustal base. The “cross-over” distance xco
between Pn and Pg is - according to Eq. (6) in IS 11.1 for shallow (near surface) sources -
roughly xco ≈ 5×zm where zm is the Moho depth. Note, however, that as focal depth increases
within the crust, xco decreases, down to about 3×zm. Accordingly, the local distance range
may vary from region to regions and range between about 100 km and 250 km. The CTBTO
Technical Instructions (see IDC Documentation,1998) considers epicentral distances between
0° to 2°, where Pg appears as the primary phase, as local distance range.
The old Manual (Willmore, 1979) defines as near earthquakes those which are observed up to
about 1000 km (or 10°) of the epicenter, and P and S phases observed beyond 10° as usually
being teleseismic phases. However, regional phases such as Pn, Sn and Lg, will generally
propagate further in stable continental regions than in tectonic or oceanic regions. According
to the Earth model IASP91, Pn may be the first arrival up to 18°. The rules published in the
IDC Documentation (1998) allow a transitional region between 17° and 20° in which phases
may be identified as either regional or teleseismic, depending on the frequency content and
other waveform characteristics. Accordingly, one might roughly define seismic sources as
local, regional and teleseismic if their epicenters are less than 2°, between 2° and 20°, or
more than 20° away from the station. Sometimes, the regional range is further subdivided into
2°-6°,where also the phase Rg may be well developed, and 6°-(17°)20° where only Sn and Lg
are strong secondary phases. However, since we have not yet found good record examples
with Pn beyond 15°, we will present and discuss our record examples for near (local and
regional) and teleseismic sources in the ranges D ≤ 15° and D > 15°, respectively.
a.) b.)
c.)
d.)
e.)
Fig. 11.39 Examples of 3-component seismograms recorded at a range of epicentral distances

from one station: a) mining-induced earthquake (D = 80 km); b) quarry blast explosion (D =
104 km); c) local earthquake (D = 110 km); d) regional earthquake (D = 504 km); and e)
teleseismic earthquake (D = 86.5°). Time scales are given below the records.
The methods used to analyze seismograms and to locate seismic sources depend on how close
they are to the recording station. For near events different programs (ORFEUS software
library; see 11.4.5) are used for source location. Differences in arrival times of phases and
slowness and azimuth estimates from the plane-wave method or frequency-wavenumber (f-k)
analysis can be used to locate distant sources with either array or network data. Time
differences between phases often give reliable distance estimates and, together with azimuth
determination from 3-component records, allow epicenters to be estimated from single station
records (see EX 11.2). If depth phases are visible and can be identified, focal depth can be
determined. Amplitude and period values of different phases are used for magnitude
estimation. Both body waves and surface waves can be used to estimate magnitude.
11.5.1 Seismograms from near sources (0° < D ≤ 15°)
Seismograms recorded at distances D ≤ 15° are dominated by P and S waves that have
traveled along different paths through the crust and the uppermost mantle of the Earth. They
are identified by special symbols for “crustal phases” (see IS 2.1). Pg and Sg, for example,
travel directly from a source in the upper or middle crust to the station whereas the phases
PmP and SmS have been reflected from, and the phases Pn and Sn critically refracted along
(or beneath) the Moho discontinuity (see Fig. 11.40). Empirical travel-time curves are given
in Exercise EX 11.1 (Figure 4) and a synthetic record section of these phases in Fig. 2.54. In
some continental regions, phases are observed which have been critically refracted from a
mid-crustal discontinuity or have their turning point in the lower crust. They are termed Pb (or
P*) for P waves and Sb (or S*), for S waves, respectively. For shallow sources, crustal
“channel-waves” Lg (for definition see IS 2.1) and surface waves Rg are observed after Sg-
waves. Rg is a short-period Rayleigh wave (T ≈ 2 s) which travels in the upper crust and is
usually well developed in records of near-surface sources out to about 300 km and thus
suitable for discriminating such events from local tectonic earthquakes (see Figs. 11.39a-c).
Fig. 11. 40 Ray traces of the main crustal

phases that are observed in the near (local and
regional) distance range from seismic sources
in a simple two-layer model of the Earth's
crust. The phase names are according to the
new IASPEI nomenclature (see IS 2.1)
(courtesy of J. Schweitzer, 2002)
.
Usually, Sg and SmS (the supercritical reflection, which often follows Sg closely at distances
beyond the critical point; see Fig. 2.40) are the strongest body wave onsets in records of near
seismic events whereas Pg and PmP (beyond the critical point) have the largest amplitudes in
the early part of the seismograms, at least up to 200 – 400 km. Note that for sub-crustal
earthquakes no reflected or critically refracted crustal phases exist. However, according to the
new IASPEI nomenclature, P and S waves from sub-crustal earthquakes with rays traveling
from there either directly or via a turning point in the uppermost mantle back to the surface
are still termed Pn and Sn (see Fig. 11.40, lower left). At larger distances such rays arrive at
the surface with apparent “sub-Moho” P and S velocities (see below).
Typical propagation velocities of Pg and Sg in continental areas are 5.5-6.2 km/s and 3.2-3.7
km/s, respectively. Note, that Pg and Sg are direct waves only to about 2° to 3°. At larger
distances the Pg-wave group may be formed by superposition of multiple P-wave
reverberations inside the whole crust (with an average group velocity around 5.8 km/s) and
the Sg-wave group by superposition of S-wave reverberations and SV to P and/or P to SV
conversions inside the whole crust. According to the new IASPEI phase nomenclature the
definitions given for Lg waves and Sg at larger distances are identical, with the addition that
the maximum energy of an Lg crustal “channel” wave travels with a group velocity around
3.5 km/s. In routine analysis, usually only the first onsets of these wave groups are picked
without noting the change in character at larger distances. According to the Technical
Instructions of the IDC Documentation (1998), stations of the CTBTO International
Monitoring System (IMS) generally tend to name the strongest transverse arrival Lg and not
Sg. A reliable discrimination is still a subject of research and not yet one of routine analysis
and data reporting. Therefore, no simple and unique criteria for discrimination, which also
depend on source type and propagation path, can be given here. They may be added to this
Manual at a later time. Lg waves may travel in continental shield regions over large distances
(see Fig. 2.15), even beyond 20° whereas Rg waves, which show clear dispersion and longer
periods than Lg (see Fig. 2.16), are more strongly attenuated and generally not observed
beyond 6°. The apparent velocities of Pn and Sn are controlled by the P- and S-wave
velocities in the upper mantle immediately below the Moho and typically range between 7.5 -
8.3 km/s and 4.4 - 4.9 km/s, respectively.
Note: Seismograms from local and regional seismic sources are strongly influenced by the
local crustal structure which differs from region to region and even between local stations.
This may give rise to the appearance of other onsets (which may be strong) between the
mentioned main crustal phases that can not be explained by a near-surface source in a single
or two-layer crustal model. Some of these phases may relate to converted waves and/or depth
phases such as sPmP (e.g., Bock et al., 1996). Also, at larger distances of up to about 30°,
multiples such as PgPg, PbPb, PnPn, PmPPmP etc. and their related S waves may be well
developed (see Fig. 11.19). However, usually these details can not be handled in routine data
analysis and epicenter location and require specialized study. For routine purposes, as a first
approximation, the IASP91 or AK135 global models (see DS 2.1) can be used for the analysis
and location of near events based on the main crustal phases. However, one should be aware
that crustal structure and velocities may differ significantly from region to region, and that
the event location can be significantly improved when local travel-time curves or crustal
models are available (see IS 11.1, Figures 11 and 12).
Fig. 11.41 shows seismograms of a shallow (h = 8 km) near earthquake from the
Vogtland/NW Bohemia region in Central Europe, recorded at seven GRSN stations in the
epicentral distance range 10 km (WERN) to 180 km (GEC2). Stations up to D = 110 km
(BRG) show only the direct crustal phases Pg, Sg, except GRFO, which in addition shows
PmP. At GEC2 Pn arrives ahead of Pg with significantly smaller amplitude. The onset times
of phases Pg, Sg and Pn were used to locate the epicenter of this event with a precision of
about 2 km. If more stations close to the epicenter are included (e.g., Fig. 11.42; D = 6 – 30
km), the precision of the hypocenter location may be in the order of a few hundred meters.
Fig. 11.41 Filtered short-period vertical component seismograms (4th order Butterworth high-
pass filter, f = 1 Hz) from a local earthquake in the Vogtland region, 04 Sept. 2000 (50.27°N,
12.42°E; Ml = 3.3). Sampling rates at the stations differ: 80 Hz for MOX, WET, CLL, and
BRG, 100 Hz for WERN and 20 Hz for GRFO and GEC2. Traces are sorted according to
epicentral distance (from 10 to 180 km). The local phases have been marked (Pg and Sg at all
stations, PmP at GRFO and Pn at the most distant station GEC2 (D = 180 km).
Fig. 11.42 Short-period recordings from stations of the local network of the Czech Academy
in Prague from a small (Ml = 3.3) local earthquake in the German/Czech border region on
June 1, 1997. The epicentral distance range is 6 km and 30 km. Such local networks allow
hypocenters to be located to better than a few hundred meters.
Fig. 11.43 (left) shows for another Vogtland swarm earthquake, a record section with
seismograms of 5 stations in the distance range 10 km to 130 km, together with the expected
travel-time curves for Pg and Sg according to an average crustal model. Fig. 11.43 (right)
shows some of the same seismograms on a map together with the station sites (triangles).
Fig. 11.43 Records of a Vogtland swarm earthquake (17 Sept. 2000; Ml = 3.1) at stations of a
local network in Germany. Left: arranged by distance together with the expected travel times
for Pg and Sg for an average crustal model; right: on a map view with station positions. The
circles indicate the position of the wavefronts of Pg (blue) and Sg (red) after 5, 10, 20 and 40
s, respectively (see also file 1 in IS 11.3 and related animation on CD-ROM).
From these two figures the following conclusions can be drawn:
• at some stations the arrival times are in good agreement with the times predicted from
an average crustal model, at other stations they are not, which implies crustal structure
varies laterally; and
• the amplitude ratio Pg/Sg varies strongly with the azimuth because of the different
radiation patterns for P and S waves. This variation can be used to derive the fault-
plane solution of the earthquake (see Figs. 3.25 and 3.26 and section 3.4.4).
Other examples of local seismograms are shown in Figs. 11.44 and 11.45. Fig. 11.44 shows
recordings from an earthquake in the Netherlands (Ml = 4.0) in the distance range 112 km to
600 km, and Fig. 11.45 those from a mining-induced earthquake in France (Ml = 3.7 ) in the
range 80 km to 500 km. These records again show obvious variation in the relative
amplitudes of Pn, Pg and Sg. The relative amplitudes depend on the distance and azimuth of
the station relative to the radiation pattern of the source, and particularly with respect to the
differences in take-off angles of the rays for the direct and the critically refracted waves (see
Fig. 11.40). The source depth with respect to the major crustal discontinuities may also
influence the relative amplitude ratio between these various phases.
Generally, for near-surface sources and distances smaller than about 400 km, Pn is much
smaller than Pg (see also Figures 3a and 3c in Datasheet 11.1). For larger distances however,
the relative amplitudes of Pn and Sn may grow so that these phases dominate the P and S
arrivals (see Fig. 2.15 and the uppermost traces in Figs. 11.44 and 11.45). This is not only
because of the stronger attenuation of the direct waves that travel mostly through the
uppermost heterogeneous crust but also because P and S near the critical angle of refraction at
the Moho form so-called “diving” phases which are not refracted into the Moho but rather
travel within the uppermost mantle with sub-Moho velocity. The recognition of these crustal
body waves and the precision of onset-time picking can be significantly improved by
stretching the time scale in digital records (compare Figures 3a and 3c in DS 11.1).
The great variability in the appearance of waveforms and relative amplitudes in near-
earthquake recordings is also illustrated by Fig. 11.46. Even seismic records at the same
station from two different sources at nearly the same distance and with similar azimuth may
look very different. This may be because the waves from the two earthquakes travel along
slightly different paths through the highly heterogeneous Alpine mountain range. However,
the fault-plane orientation and related energy radiation with respect to the different take-off
angles of Pn and Pg, may also have been different for these two earthquakes.
Fig. 11.44 Vertical-component short-period filtered broadband seismograms (4th order

Butterworth high-pass filter, f = 0.7 Hz; normalized amplitudes) from a local earthquake at
Kerkrade, Netherlands, recorded at 15 GRSN, GRF, GERES and GEOFON stations. Ml =
4.0; epicentral distances between 112 km (BUG) and 600 km (GEC2). Note the variability of
waveforms and relative phase amplitudes of local/regional earthquakes in network recordings
in different azimuths and epicentral distances. The suitability of filters for determination of
local phase onsets has to be tested. Local magnitudes determined from a Wood-Anderson
simulation.
Fig. 11.45 Vertical-component short-period filtered BB seismograms (4th order Butterworth
high-pass filter, f = 0.7 Hz, normalized amplitudes) from a local mining-induced earthquake
at the French-German border recorded at 11 GRSN, GERES and GEOFON stations (Ml =
3.7; epicentral distances between 80 km (WLF ) and 501 km (GEC2).
Note that in Figs. 11.39a and b and Fig. 11.45 (e.g., station WLF), the longer period Rg
waves, following Sg, are particularly well developed in records of near-surface quarry blasts
or shallow mining-induced earthquakes but not in the natural earthquake records (as in Fig.
11.43).
As mentioned above, at distances beyond about 600 to 800 km, Pn and Sn become the
dominating body-wave onsets that for shallow sources are followed by well-developed
surface-wave trains. Figure 6a in DS 11.1 shows a typical 3-component BB-velocity record of
such an earthquake in Italy made at station GRA1 in Germany (D = 10.3°). Figure 6c shows
the respective BB recordings of the same shock at 10 stations of the GRSN (D = 8°- 12°). Pg
and Sg are no longer recognizable. In fact, Pn and Sn at these regional distances are no longer
pure head waves from the Moho discontinuity but rather so-called diving phases of P and S
which have penetrated into the uppermost mantle but travel also with the sub-Moho velocity
of Pn of about 8 km/s. These diving phases may be of longer periods than Pn at shorter
distances. One should also be aware that local and regional earthquakes do not only appear in
short-period recordings. Strong near events with magnitudes above 4 usually generate also
strong long-period waves (see Figs. 11.8 and 11.10).
Pn Pg Sn Sg
Fig. 11.46 Comparison of Z-component short-period filtered records at station MOX,

Germany, of two earthquakes in Northern Italy ( trace 1: 28 May 1998; trace 2: 24 Oct. 1994)
at about the same epicentral distance (D = 505 km and 506 km, respectively) and with only
slightly different backazimuth (BAZ = 171° and 189°, respectively). Note the very different
relative amplitudes between Pn and Pg, due to either crustal heterogeneities along the ray
paths or differences in rupture orientation with respect to the different take-off angles of Pn
and Pg rays.
In general, regional stations and local networks complement each other in the analysis of smaller sources at local
distances. Additionally, source processes and source parameters can be estimated using local station data. For
this purpose, first motion polarities (compression c or +, dilatation d or -) for phases Pg, Pn, Sg and amplitude
ratios (P/SV) should be measured for fault plane solution and moment tensor inversion (see 3.4 and 3.5). In
regions with a poor station coverage, the mean precision of location may be several
kilometers and source depths may then only be determined with teleseismic depth phases by
way of waveform modeling (see 2.8).
An important aspect to consider in digital recordings and data analysis of local and regional
seismograms is the sampling rate. Sampling with more than 80 s.p.s. is generally suitable for
near seismic events. With lower sampling rate ssome of the most essential information about
the seismic source process such as the corner frequency of the spectrum and its high-
frequency decay, may be lost. Fig. 11.47 gives an example.
Fig. 11.47 Top: BB-velocity seismogram from a local earthquake near Bad Ems (11 Oct.
1998; Ml = 3.2) recorded at the GRSN station TNS (D = 40 km). Different sampling rates
were used for data acquisition. Traces 1 – 3 were sampled at 20 Hz and traces 4 – 6 at 80 Hz.
In the records with the higher sampling rate, the waveforms are much more complex and
contain higher frequencies. The high frequency content is suppressed with the lower
sampling rate. Bottom: Fourier spectrum of traces 1 (sampling rate 20 Hz – pink) and 4
(sampling rate 80 Hz – red). The lower sampling rate cuts off the high-frequency components
of the seismic signal. Thus the corner frequency of the signal at about 20Hz and the high-
frequency decay could not be determined from the pink spectrum. The green spectrum
represents the seismic noise.
In the example considered in Fig. 11.47 only the 80 Hz data stream with a Nyquist frequency
of 40 Hz allows the corner frequency near 20 Hz to be determined. However, in some regions,
or when studying very small local earthquakes, still higher frequencies have to be analyzed.
This may require sampling rates between 100 and 250 Hz.
Note, that besides the regional phases Sn and Lg the teleseismic phase PcP also may be
observed in the far regional distance range (6°-20°) in short-period seismograms of strong
events (Ml > 4). PcP, which gives a good control of source depth, can be identified in array
recordings because of its very small slowness.
In Box 1 below a summary is given of essential features that can be observed in records of
local and regional seismograms. For more records in this distance range see DS 11.1.
BOX 1: General rules for local and regional events
• The frequency content of local events (D < 2°) is usually high (f ≈ 0.2 - 100 Hz).
Therefore they are best recorded on SP or SP-filtered BB instruments with
sampling rates f ≥ 80 Hz. The overall duration of short-period local and regional
(D < 20o) seismograms ranges between a few seconds and to several minutes.
• Strong local/regional sources radiate long-period energy too and are well
recorded by BB and LP seismographs. In the far regional range the record
duration may exceed half an hour (see Fig. 1.2).
• Important seismic phases in seismograms of local sources are Pg, Sg, Lg and Rg
and in seismograms of regional sources additionally Pn and Sn, which arrive
beyond 1.3°-2° as the first P- and S-wave onsets. The P waves are usually best
recorded on vertical and the S waves on horizontal components.
• Note that Pg is not generally seen in records from sources in the oceanic crust.
Also, deep (sub-crustal) earthquakes lack local and regional crustal phases.
• For rough estimates of the epicentral distance D [km] of local sources, multiply
the time difference Sg-Pg [s] by 8, and in the case of regional sources the time
difference Sn-Pn [s] by 10. For more accurate estimates of D use local and
regional travel-time curves or tables or calculations based on more appropriate
local/regional crustal models.
• The largest amplitudes in records of local and regional events are usually the
crustal channel waves Lg (sometimes even beyond 15°), and for near-surface
sources the short-period fundamental Rayleigh mode Rg. For near-surface
explosions or mining-induced earthquakes, Rg, with longer periods than Sg, may
dominate the record, however usually not beyond 4°.
• For routine analysis the following station/network readings should be
made: (1) the onset time and polarity of observed first motion phases; (2)
onset times of secondary local and regional phases; (3) local magnitude
based either on maximum amplitude or duration. If local/regional
calibration functions, properly scaled to the original magnitude definition
by Richter (1935), are not available it is recommended to use the original
Richter equation and calibration function, together with local station
corrections.
11.5.2 Teleseismic earthquakes (15° < D < 180°)
11.5.2.1 Distance range 15° < D ≤ 28°
At distances beyond 15°, sometimes referred to as “far-regional”, Pn and Sn amplitudes

become too small (except in some shield regions) and the first arrival phase has a travel path
through deeper parts of the upper mantle. The common nomenclature for these waves with
longer periods than Pn and Sn is P and S, respectively. For the next 12°, the records look
rather simple in one respect, namely, that only three major wave groups are recognizable (P,
S, and surface waves, see Fig. 11.48).
Fig. 11.48 Three-component broadband seismograms from an earthquake in Iceland on 04

June 1998, recorded at GRF-array station GRA1 (D = 22.5°). P, S and surface waves are
recognizable. Horizontal traces have been rotated (R = Radial, T = Transverse) with R
showing into the source direction. The overall duration of the record is about 15 min.
The body-wave groups themselves, however, are rather complicated because of the refraction
and reflection of P and S at the pronounced velocity increase at the bottom of the upper
mantle (410 km discontinuity) and at the bottom of the transition zone to the lower mantle
(660 km discontinuity). These strong increases/gradients in wave velocity give rise to the
development of two triplications of the P-wave travel-time curve with prograde and
retrograde branches which in some distance ranges follow closely each other (see Fig. 2.29),
thus forming a sequence of successive P- (and S-) wave onsets (see Fig. 11.49 ). The largest
amplitudes occur in the range of the left-side cusp of the 660 km discontinuity triplication
(P660P) between about 18° and 20° (also termed “20°-discontinuity”) but with weaker P-
wave first arrivals some 5 to 10 s earlier. Accordingly, small differences in epicentral distance
can lead to large differences in the appearance of the body-wave groups in seismic records
(see Fig. 11.49). Generally, P waves are rather complex and wave onsets emergent. Surface
waves of shallow earthquakes, however, are strong, clearly separated from S waves and very
useful for surface wave magnitude (Ms) estimation.
Fig. 11.49 shows BB-velocity seismograms from an earthquake in Turkey recorded at stations
of the GRSN in the distance range between about 14.5° and 19.5°. Note the increase of the P-
wave amplitudes with distance when approaching “20°-discontinuity”.
Fig. 11.49 Vertical-component BB-velocity seismograms from the damaging earthquake in

Düzce, NW-Turkey, recorded at 12 GRSN-stations in the distance range 14.5° to 19.5°.
(Source data from NEIC-QED: 12 Nov. 1999, OT 16:57:20; 40.79°N, 31.11°E; h = 10 km;
Mw = 7.1; D = 16.5° and BAZ = 115° from GRFO). Traces are sorted according to distance.
Incoherent and complex P waves are followed by weak S waves and distinct and clearly
dispersed surface waves, which have longer periods than the S waves. The body waves P and
S are affected by upper mantle discontinuities. Note the increase in P-wave amplitudes with
distance due to the cusp of P660P around 18° to 19° (see Fig. 2.29).
Fig. 11.50 shows 3-component records (Z, R, T) demonstrating the presence of the 20°
discontinuity in another part of the world. The seismograms are from an earthquake that
occurred in the southern part of the New Hebrides at 35 km depth, recorded between 16.9°
and 23.6° by a regional network of portable broadband instruments deployed in Queensland,
Australia (seismometers CMG3ESP; unfiltered velocity response). The complex P wavelets
appear in Z and R only. In their later parts they may contain PnPn arrivals. Similarly complex
S-wave groups appear in R and T, and may include SnSn.
Fig. 11.50 Three-component BB records made in Queensland, Australia, from an earthquake

in the New Hebrides between D = 16.9° and 23.6°. Note the complexity of P- and S-wave
arrivals around the 20° discontinuity. On each set of records the predicted phase arrival times
for the AK135 model (see Fig. 2.53) are shown as faint lines. However, there might be no
clear onset visible at these times above the noise or signal-coda level of previous phases.
Never use such theoretical onset marks for picking onset times! (Courtesy of B.L.N. Kennett).
11.5.2.2 Distance range 28° < D ≤ 100°
The main arrivals at this distance range up to about 80°, have traveled through the lower
mantle and may include reflections from the core-mantle boundary (CMB) (Fig. 11.53). The
lower mantle is more homogeneous than the upper mantle (see Fig. 2.53). Accordingly, P and
S waves and their multiples form rather simple long-period seismograms (Figs. 11.9 and
11.52; see also files 4 and 5 in IS 11.3 and animation on CD-ROM). Between 30° and 55°, the
waves reflected from the core (e.g., PcP, ScP etc.) are also often recorded as sharp pulses on
short-period records, particularly on records of deep earthquakes where depth phases appear
well after the core reflections (see Fig. 11.16). At around 40°, the travel-time curve of PcP
intersects those of PP and PPP (see Fig. 11.53) and in horizontal components PcS intersects S,
and ScS intersects SS and SSS. This complicates proper phase separation, at least for the later
phases on long-period records, where SS and SSS may be strong. ScP, however, may also be
rather strong on short-period vertical components (see Fig. 11.53). Note that PP, PS, SP and
SS are Hilbert-transformed (see 2.5.4.3). Their onset and amplitude picks can be improved by
inverse Hilbert transformation, which is part of modern analysis software such as Seismic
Handler (SH and SHM). The amplitudes of the core reflections decrease for larger distances
but they may be observed up to epicentral distance of about 80° (ScP and ScS) or 90° (PcP),
respectively, beyond which ScS merges with the travel-time curves of SKS and S and PcP
with that of P (compare travel-time curves in Figure 4 of EX 11.2 with Figs. 11.16 and 11.55).
Fig. 11.51 Seismic ray paths through the mantle and core of the Earth with the respective
phase names according to the international nomenclature (see Fig. 2.48 and overlay for related
travel-time curves, and IS 2.1 for phase names and their definition). The red rays relate to the
3-component analog Kirnos SKD BB-displacement record of body waves from an earthquake
at D = 112.5° at station MOX, Germany.
Fig. 11.52 Long-period Z- (left) and T-component seismograms (middle) of a shallow

earthquake in western India recorded in the distance range 51° to 56° at stations of the GRSN.
Two cutout sections from short-period Z-component records of multiple core phases are
shown on the right and the related ray paths at the top (for animation see file 4 in IS 11.3).
Fig. 11.53 Vertical-component Kirnos SKD BB-displacement (left) and WWSSN-SP

seismograms (right) from an intermediate depth (h = 227 km) earthquake in the Afghanistan-
Tajikistan border region recorded at stations of the GRSN. Besides P the depth phases pP, sP,
pPP, sPP and pPPP and the core reflections PcP and ScP are clearly visible, particularly on
the short-period records. The ray traces of these phases are shown in the upper right corner
(see also file 3 in IS 11.3 and related animation on CD-ROM).
Fig. 11.54 shows the ray paths for S, ScS and SKS and their related travel-time curves
according to the IASP91 model for the whole distance range from 60° to 180° and Fig. 11.55
both short- and long-period records for these waves between 50° and 80°.
Fig. 11.54 Ray paths for S, ScS and SKS and their related travel-time curves according to the
IASP91 model for the whole distance range from 60° to 180°.
Fig. 11.55 SP (left) and LP (right) horizontal-component seismograms from a deep-focus

earthquake in the Sea of Okhotsk (20 April 1984, mb = 5.9, h = 588 km) recorded by stations
in the distance range 50.1° to 82.2°. Note the different amplitude scaling. Accordingly, the
amplitudes of various transverse phases are 2 to 10 times larger in long-period records when
compared with short-period records. Four distinct phases are identified: S, ScS, sS and SS.
SKS, which emerges at distances larger than 60° however, overlaps with ScS between about
65° and 75°. S, ScS and SKS start to coalesce as distance increases toward 82°. Beyond this
distance SKS arrives before S, SKKS and ScS (reprinted from Anatomy of Seismograms,
Kulhánek, Plate 40, p.137-138;  1990; with permission from Elsevier Science).
Arrays and network records, which also allow f-k and vespagram analysis are very useful for
identifying the core reflections PcP, ScP and ScS because their slownesses differ significantly
from those of P, S and their multiple reflections (see Fig. 11.52 as well as Figures 6a and b
and 7b in DS 11.2). Surface reflections PP, PPP, SS and SSS are well developed in this
distance range in long-period filtered records and converted waves PS/SP at distances above
40°. Sometimes the surface reflections are the strongest body-wave onsets at large distances
(see Figures 10c and 11b in DS 11.2). Their identification can be made easier when network
records are available so vespagram analysis can be used (e.g., Figure 11c in DS 11.2). In
short-period filtered network records it is sometimes also possible to correlate well in this
distance range multiple reflected core phases such as PKPPKP or P'P', SKPPKP and even
SKPPKPPKP (see Fig. 1.4).
Beyond 83° SKS moves ahead of S and its amplitude relative to S increases with distance.
Network and array analysis yields different slowness values for S and SKS because of their
diverging travel-time curves (see Fig. 11.54). This helps to identify these phases correctly.
Note that the differential travel time SKS-P increases only slowly with distance (see Figure 4
in EX 11.2). Misinterpretation of SKS as S may therefore result in an underestimation of D by
up to 20°! Since SKS is polarized in the vertical plane it can be observed and separated well
from S in radial and vertical components of rotated seismograms (see Figures 10c, 13e, 14e,
and 15b in DS 11.2). The same applies for PcS and ScP, which are also polarized in the
vertical plane in the direction of wave propagation.
In the distance range between about 30° and 105° multiple reflected core phases P'N or
between about 10° < D° < 130° the phases PNKP, with N-1 reflections either at the free
surface (P'N) or from the inner side of the core-mantle boundary (PNKP) may appear in short-
period records some 13 min to 80 min after P. An example for PKPPKP (P'P') and PKKKKP
(P4KP) is given in Fig. 11.52. These phases are particularly strong near caustics, e.g., P'P'
(see Fig. 11.69) and P'P'P' (P'3) near 70° and PKKP near 100° (see Fig. 11.71) but they are
not necessarily observable at all theoretically allowed distances. Figures 9 and 10 in EX 11.3,
however, document the rather wide distance range of real observations of these phases at
station CLL (for P'P' between 40° and 105° and for PKKP between 80° and 126°). Note the
different, sometimes negative slowness of these phases. More record examples, together with
the ray paths of these waves, are presented in a special section on late core phases (11.5.3).
For differential travel-time curves PKKP-P and PKPPKP-P see Figures 9 and 10 in EX 11.3.
Also PKiKP, a weak core phase reflected from the surface of the inner core (ICB), may be
found in short-period array recordings throughout the whole distance range, about 4.5 to 12
min after P. Its slowness is less than 2s/°. Beyond 95°, P waves show regionally variable,
fluctuating amplitudes. Their short-period amplitudes decay rapidly (see Fig. 3.13) because of
the influence of the core (core-shadow) while long-period P waves may be diffracted around
the curved core-mantle boundary (Pdif, see Figs. 11.59 and 11.63 as well as Figures 1, 2, 4b
and 6c and b in DS 11.3).
In any event, comprehensive seismogram analysis should be carried out for strong
earthquakes which produce many secondary phases. Unknown phase arrivals should also be
reported for further investigations into the structure of the Earth. When reporting both
identified and unknown phases to international data centers the IASPEI-proposed
international nomenclature should strictly be observed (see IS 2.1).
11.5.2.3 Distance range 100°< D ≤ 144°
Within this distance range, the ray paths of the P waves pass through the core of the Earth.
Due to the large reduction of the P-wave velocity at the core-mantle boundary (CMB) from
about 13.7 km/s to 8.0 km/s (see Fig. 2.53) seismic rays are strongly refracted into the core
(i.e., towards the normal at this discontinuity). This causes the formation of a "core shadow".
This "shadow zone" commences at an epicentral distance around 100°. The shadow edge is
quite sharp for short-period P waves but diffuse for long-period P and S waves that are
diffracted around the curved CMB (compare Figures 6b and c as well as 7a and b in DS 11.3).
For strong earthquakes Pdif and Sdif may be observed out to distances of about 150° (see
Figs. 11.56, 11.59 and 11.63).
Fig. 11.56 SRO-LP filtered 3-component seismograms at station GRA1, Germany, in D =

117.5° from an earthquake in Papua New Guinea (17 July 1998, Ms = 7.0). The N and E
components have been rotated into the R and T directions. Phases Pdif, PP, PPP (not marked)
and a strong SP are visible on the vertical component, whereas the phases SKS, SKKS and
PS, which are polarized in the vertical propagation plane, are strong on the radial (R)
component (as are PP and PPP). Sdif, SS and SSS are strong on the transverse (T) component.
Note that Pdiff and Sdiff are still acceptable alternative phase names for Pdif and Sdif.
Fig. 11.57 shows rotated (R-T) horizontal component SRO-LP recordings at GRSN stations
from two intermediate deep events in the Chile-Bolivia border region and in the Mariana
Islands, respectively. The related ray paths are depicted in the upper part. The records cover
the transition from the P-wave range into the P-wave core shadow. Magnified cut-outs, also of
the related Z-component records, are presented for both earthquakes in Figures 1 and 2 of DS
11.3. They show more clearly the first arriving longitudinal waves and their depth phases. The
following conclusions can be drawn from a comparison of these figures:
• Pdif arrives about 4 minutes (at larger distance up to 6 min; see Fig. 11.63) earlier
than the stronger PP;
• The largest phases (see also Fig. 11.60) are usually PP, PPP, PS, SP, Sdif, SKS,
SKKS, SS and SSS;
• SKS is the first arriving shear wave, followed by SKKS, SP or PS (and the related
depth phases), all on the R component;
• S/Sdiff and SS may be strong(est) in T or R, or even in both components, depending
on the SV/SH ratio of shear-wave energy radiated by the source.
Fig. 11.57 SRO-LP filtered records of GRSN stations on R and T components in the distance
range between 96° and 109° from two earthquakes on opposite backazimuth. Left: Chile-
Bolivia Border region; right: Mariana Islands (source data according to the NEIC).
If no Pdif is observed, PKiKP is the first arrival in short-period records up to 113° (see Figure
3b in DS 11.3). For distances beyond 114° PKiKP follows closely after PKPdf (alternatively
termed PKIKP). The latter has traveled through the outer and inner core and arrives as first
onset for D ≥ 114°. PKPdf is well recorded in short-period seismograms but usually with
emergent onsets and, up to about 135° distance, still with weaker amplitudes than PKiKP. Fig.
11.58 shows the amplitude-distance relationship between PKiKP, PKPdf and the other direct
core phases PKPab and PKPbc, which appear with largest amplitudes beyond 143°. Fig. 11.59
depicts the ray paths and travel-time curves of Pdif, PKiKP, PKPdf, PKPab and PKPbc (for
more complete ray paths see IS 2.1). Also PKKP (with its branches ab and bc) is often clearly
recorded between 110° and 125° (see Figures 3c and d in DS11.3).
Fig. 11.58 Smoothed amplitude-distance relationships for the core phases PKiKP, PKPdf,
PKPab and PKPbc as calculated for the model 1066B in the distance range 130° to 160°
(modified from Houard et al., Amplitudes of core waves near the PKP caustic,…Bull. Seism.
Soc. Am., Vol. 83, No. 6, Fig. 4, p. 1840,  1993; with permission of Seismological Society
of America).
Fig. 11.60 presents records of GRSN stations in the distance range 121° to 127° from an
earthquake of intermediate depth (h = 138 km) in the region of New Britain (see file 7 in IS
11.3 and animation CD). They show the PKPdf arrivals about 3.5 min after Pdif together with
the dominant phases in this range, namely PP, PPP, PS, PPS and the Rayleigh-wave arrival
LR in the Z component and the SS, SSS and the Love-wave arrival LQ in the T component.
Also shown, together with the ray paths, is the core phase P4KPbc, which has been reflected 3
times at the surface of the Earth, and which is recognizable only on the short-period filtered
vertical component. Between about 128° and 144° some incoherent waves, probably scattered
energy from “bumps” at the CMB, may arrive as weak forerunners up to a few seconds before
PKPdf. They are termed PKPpre (old PKhKP). PKPdf is followed by clear PP, after about 2
to 3 minutes, with SKP or PKS arriving about another minute later (see Fig. 11.61).
SKP/PKS have their caustics at about 132° and thus, near that distance, usually have rather
large amplitudes in the early part of short-period seismograms (see Fig. 11.61). For medium-
sized earthquakes these phases may even be the first ones to be recognized in the record and
be mistaken for PKP. Note that for near-surface events SKP and PKS have the same travel
time, but with the former having relatively larger amplitudes in the Z component whereas
PKS is larger in the R component. For earthquakes at depth, PKS and SKP separate with the
latter arriving earlier the deeper the source (Fig. 11.61). Beyond 135° there are usually no
clear phases between SKP and SS. Misinterpretation (when Pdif is weak or missing) of PP
and SKS or PS waves as P and S may in this distance range result in strong underestimation
of D (up to 70°). This can be avoided by looking for multiple S arrivals (SS, SSS) and for
surface waves which follow more than 40 min later (see Table 5 in DS 3.1). For more records
see DS 11.3, examples 1 to 7.
Fig. 11.59 Ray paths of Pdif, PKPdf, PKPab and PKPbc and their travel-time curves for
surface focus and deep focus (h = 600 km) events.
Fig. 11.60 Main seismic phases in the distance range 121° to 127° on records made at GRSN
stations. Left and middle: SRO-LP filtered Z and T component, respectively. Right: SRO-LP
and WWSSN-SP components. Top right: Ray traces of phases shown (see also file 7 in IS
11.3 and animation on CD-ROM).
Fig. 11.61 Vertical-component seismograms at GRSN stations recorded in the distance range
135° to 141°. Left: Kirnos SKD BB-displacement; right: WWSSN-SP; top: ray paths of the
phases PKPdf , PP and SKP/PKS bc. Note the precursor PKPpre.
11.5.2.4 Core distance range beyond 144°
Between 130° and 143° the first onsets of longitudinal core phases are relatively weak and
complex in short-period records, but their amplitudes increase strongly towards the caustic
around 144°. At this epicentral distance three PKP waves, which have traveled along different
ray paths through the outer and inner core, namely PKPdf, PKPbc (old PKP1) and PKPab (old
PKP2) arrive at the same time (see Fig. 11.59) so their energies superimpose to give a strong
arrival with amplitudes comparable to those of direct P waves at epicentral distances around
40° (compare with Fig. 3.13). Beyond the caustic the travel-time curves of these three PKP
waves split into the branches AB (or ab), BC (or bc) and DF (or df) (see Fig. 11.59).
Accordingly, the various arrivals can be identified uniquely by attaching to the PKP symbol
for a direct longitudinal core phase the respective branch symbol (see Figs. 11.59 and 11.62).
Note that the PKPbc branch shown in Figs. 11.59 between the point B and C is ray-
theoretically not defined beyond 155°. However, in real seismograms one often observes
weak onsets between PKPdf and PKPab up to about 160° or even slightly beyond in the
continuation of the PKPbc travel-time curve. This phase is a PKP wave diffracted around the
inner core boundary (ICB) and named PKPdif (see Fig. 11.62 and 11.63).
Fig. 11.62 Left: Records of the direct core phases PKPdf, PKPbc and PKPab as well as of
the diffracted phase PKPdif from a Kermadec Island earthquake at stations of the GRSN in
the distance range between 153° and 159°; right: ray paths through the Earth.
The relative amplitudes between the three direct longitudinal core phases change with
distance. In SP records these three phases are well separated beyond 146°, and PKPbc is the
dominant one up to about 153° though the separation between these three phases is not clear
within this range in LP records (Fig. 11.63 upper part and Figures 9a-c in DS 11.3).
Fig. 11.63 Short-period (bottom) and long-period (SRO-LP) filtered broadband records of
GRSN stations of PKP phases in the distance range 116° to 163°. In the LP records
additionally the onsets of Pdif and PP have been correlated with their travel-time curves.
On short-period records the phases PKPdf, PKPbc and PKPab are easy to identify on the basis
of their typical amplitude and travel-time pattern. D can be determined with a precision better
than 1.5° by using differential travel-time curves for the different PKP branches (see EX
11.3). On records of weaker sources, PKPbc is often the first visible onset because the PKPdf,
which precedes the PKPbc, is then too weak to be observed above noise. On long-period
records superposed different onsets may be recognizable only at distances larger than 153°.
Then PKPab begins to dominate the PKP-wave group on short-period records (compare Figs.
11.63). Towards the antipode, however, PKPdf (PKIKP) becomes dominating again whereas
PKPab disappears beyond 176°.
On LP and BB records the dominant phases on vertical and radial components are PKP, PP,
PPP and PPS while on the transverse component SS and SSS are dominant. For deep sources,
their depth phases sSS and sSSS may be strongest (see Fig. 11.64). Besides PP, which has
traveled along the minor arc (epicentral distance D) the phase PP2, which has taken the longer
arc to the station (360° - D), may be observable, as well as phases such as PcPPKP and others
(see Fig. 11.65 as well as file 9 in IS 11.3 and related animation on CD-ROM). SKKS,
SKKKS, SKSP etc. may still be well developed on radial component records (see also Figs.
2.48 and 2.49 with the related travel-time curve overlay). The whole length of BB or LP
seismograms in this distance range between the first onsets and the surface wave maximum is
more than an hour (see Tab. 5 in DS 3.1).
11.5.3 Late and very late core phases

For large magnitude sources, reflected core phases may be observed in addition to the direct
ones, sometimes with up to 4 (or even more) repeated reflections. These phases may be observed
at practically all teleseismic distances with delays behind the first arriving P or PKP onsets
ranging from about 10 minutes up to about 80 minutes, depending on the number of multiple
reflections. These phases are clearly discernible only in high-magnifying SP (or appropriately
filtered BB) records. Most frequently observable are the single surface reflection P'P' (also
termed PKPPKP), and the single reflection from the inner side of the core-mantle boundary,
PKKP. As for the direct core phases, these multiple reflections develop different travel-time
branches according to their different penetration depth into the outer core (see also figures in IS
2.1). Figs. 11.66 and 11.67 show the ray paths for P'P' and PKKP waves, respectively, together
with their related IASP91 travel-time curves (Kennett and Engdahl, 1991) for sources at depth h
= 0 km and h = 600 km. Where there is more than one reflection the respective phases are often
written P'N or PNKP, respectively, with the number of reflections being N-1. Ray paths and
short-period record examples for P'N with N = 2 to 4 are shown in Figs. 11. 68 to 11.70 and for
PNKP with N = 2 to 5 in Figs. 11.70 and 71. Fig. 11.64 shows a P5KP (PKKKKKP), which has
been four times reflected from the inner side of the CMB. It is observed nearly 37 min after
PKP. The phase P7KP has been found in a record at Jamestown, USA, of an underground
nuclear explosion on Novaya Zemly in 1970.
All these figures show that these late arrivals may still have a significant SNR. Since they appear
very late and thus isolated in short-period records, station operators may wrongly interpret them
as being P or PKP first arrivals from independent events. This may give rise to wrong phase
associations and event locations, which, particularly in a region of low seismicity, may give a
seriously distorted picture of its seismicity. This was demonstrated by Ambraseys and Adams
(1986) for West Africa.
Fig. 11.64 Records of GRSN stations of a deep earthquake in the Fiji Island region. Top: ray
paths and source data; bottom: records on the Z component (LP left and SP right) and T
component (middle) (see also file 8 in IS 11.3 and related animation on CD-ROM).
Fig. 11.65 Records of GRSN stations of a shallow (crustal) earthquake east of the North
Island of New Zealand. Upper right: ray paths, source data and wavefronts of PP and PP2
arriving in Germany; bottom: records on long-period components (Z left and right; T middle).
An animation has been produced that shows the ray propagation and seismogram formation
for this earthquake (see file 9 in IS 11.3 and related CD-ROM).
Fig. 11.66 Ray paths and travel-time curves for P'P' according to the Earth model IASP91
(Kennett and Engdahl, 1991).
Fig. 11.67 Ray path and travel-time curves for PKKP according to the Earth model IASP91
for epicentral distances of the source between 60o and 140o.
Fig. 11.68 Ray path (top) and short-period Z-component records at stations of the GRSN
(bottom) of P, P'P' and P'3 together with their theoretically expected arrival times according to
the IASP91 and JB tables. Earthquake in Myanmar; distance range 65° < D < 70°.
Fig. 11.69 Short-period record segments showing P and PKPdf (bottom) together with P'2,
P'3 and P' 4 (middle) at GRSN stations. Top: Related ray paths and source data. Note the
negative slowness for P'2 and P'4. The theoretical travel-time curves relate to IASP91. Record
length is one minute.
Fig. 11.70 Late and very late multiple core phases PKKP, P'2 and P'3, respectively, together
with their depth phases in short-period filtered record segments of GRSN stations from an
earthquake in Northern Peru at an epicentral distance around 92°. For an animation of ray
propagation and seismogram formation from this source see file 6 in IS 11.3 and related CD.
Fig. 11.71 Short-period record segments of P and PKPdf (bottom) together with those of
PKKP, P3KP and P5KP (middle) at GRSN stations. Top: Related ray paths and source data.
Note the negative slowness for PKKP and P3KP. The theoretical travel-time curves relate to
IASP91. Record length is one minute.
In Figs. 11.62 to 11.71 the theoretical travel-time curves for core phases have been shown.
For consistency, they are based on the travel-time model IASP91 (Kennett and Engdahl,
1991), as in all earlier record sections shown for the teleseismic distance range. An exception
is Fig. 11.68, which shows additionally the theoretical travel-time curve for the JB model
(Jeffreys and Bullen , 1940). One recognizes, that the model IASP91 yields onset times for
core phases that tend to be earlier than the real onsets in the seismograms. This applies to both
direct and multiple reflected core phases. The agreement between real and theoretical onsets
of core phases is better when using the JB model. The JB model is still regularly used for the
location of teleseismic sources at the international data centers in Boulder (NEIC), Thatcham
(ISC) and Moscow whereas the IDC of the CTBTO uses IASP91. The more recent model
AK135 (Kennett et al., 1995) is more appropriate than IASP91 for core phases. No
recommendations have been made yet by IASPEI for using a best fitting global 1-D Earth
model as standard at all international data centers. However, the NEIC is currently rewriting
its processing software so that it will allow to use different Earth models, and AK135 will
probably be its “default” model.
Note that the difference between the azimuth of the P wave and that of P'P' and PKKP,
respectively, is 180° (see Figs. 11.66 and 11.67). The related angular difference of the surface
projections of their ray paths is 360°–D where D is the epicentral distance. Accordingly, the
slowness of P'P' as well as of any even number P'N is negative, i.e. their travel time decreases
with D. This also applies to PKKP and P3KP, as can be seen from Fig. 11.71. The surface
projection of the travel paths of P'3 is 360° + D and that of P'4 is 2 × 360°–D. PKPPKP is
well observed between about 40° < D ≤ 105°. In this range it follows the onset of P by 33 to
24 min (see Figure 10 in EX 11.3 with observed data). The existence of P'N is not limited to
PKPbc. Fig. 11.68 shows an example of P'3df, recorded at a distance of about 67°. P'4 is
sometimes observed in the distance range 112° to 136°. An example is given in Fig. 11.69.
Similar ray paths can be constructed for PNKP, the phase with (N-1) reflections from the
inner side of the core-mantle boundary (see Fig. 11.71). Figure 9 in EX 11.3 gives the
differential travel-time curves for PKKP to the first arrivals P or PKP, respectively, in the
distance range between 80° and 130° together with the observed data. In this range PKKP
arrives 13 to 19 min behind P or 9.5 to 12 min behind PKPdf. Higher multiple reflections from
the inner side of the CMB such as P3KP, P4KP and P5KP are observed, if at all, at 37 ± 1 min
after the first arriving wave. The latter is true for P3KP following P at around 10°, for P4KP
following P between 45° < D < 75° and for P5KP following the onset of PKPdf between about
130° < D < 150° (called “37- min” rule- of- thumb).
A particular advantage of these multiple reflected core phases is the small depth dependence
of their travel-time differences to P and PKP, respectively. Consequently, their identification
allows very good distance estimates to be made from single station records even when the
source depth is not known. Because of the inverse differential travel-time curves of PKPPKP
and PKKP with respect to P and PKP their identification can be facilitated by comparing the
onset times at neighboring stations (e.g., Fig. 11.70). The polarization of both the first arrival
and the possible PKKP or PKPPKP onset, determined from 3-component records, can also aid
identification because their azimuths should be opposite to that of P or PKIKP, respectively.
Sometimes, also converted core reflections such as SKPPKP or SKKP can be observed in
short-period recordings. However, direct or reflected core phases, which have traveled along
both ray segments through the mantel as S waves (such as SKS, SKKS, etc.) are mostly
observed in broadband or long-period records.
11.5.4 Final remarks on the recording and analysis of teleseismic events
Box 2 below, summarizes the key criteria that should be taken into account when recording
and analyzing seismograms from sources at teleseismic distances (see also 11.2.6.1).
Box 2: General rules for recording and analyzing teleseismic events
• The overall duration of teleseismic records at epicentral distances larger than

15o(or 20°) ranges from tens of minutes to several hours. It increases both
with epicental distance and the magnitude of the source.
• High frequencies, of S waves in particular, are attenuated with distance so
recordings at long range are generally of lower (f ≈ 0.01 - 1 Hz) frequency
than local or regional recordings.
• Usually only longitudinal waves, both direct or multiple reflected P and PKP
phases, which are much less attenuated than S waves, are well recorded by
short-period, narrow-band seismographs (or their simulated equivalents) with
high magnification of frequencies around 1 Hz. However, S waves from deep
earthquakes may sometimes be found also in SP teleseismic records.
• Because of the specific polarization properties of teleseismic body and
surface waves, polarization analysis is an important tool for identifying the
different types of wave arrivals.
• According to the above, teleseismic events are best recorded by high-
resolution 3-component broadband seismographs with large dynamic range
and with sampling rates f = 20 Hz.
• The main types of seismic phases from teleseismic sources are (depending
on distance range) the longitudinal waves P, Pdif, PKP, PcP, ScP, PP, and
PPP and the shear waves S, Sdif, SKS, ScS, PS, SS, and SSS. The
longitudinal waves are best recorded on vertical and radial components
whereas the shear waves appear best on transverse and/or radial components.
• Multiple reflected core phases such as P'N and PNKP, which appear on SP
records as isolated wavelets, well separated from P or PKP, may easily be
misinterpreted as P or PKP arrivals from independent seismic sources if no
slowness data from arrays or networks are available. Their proper
identification and careful analysis helps to avoid wrong source association,
improves epicenter location and provides useful data for the investigation of
the deeper interior of the Earth.
• Several body wave phases such as PP, PS, SP, SS, PKPab and its depth
phases, SKKSac, SKKSdf, PKPPKPab, SKSSKSab undergo phase shifts and
wavelet distortions at internal caustics (see 2.5.4.3). This reduces the
accuracy of their time and amplitude picks and their suitability for improving
source location by waveform matching with undistorted phases. Therefore it
is recommended that seismological observatories correct these phase shifts
prior to parameter readings by applying the inverse Hilbert-transformation,
which is available in modern software for seismogram analysis.
• Surface waves of shallow events have by far the largest amplitudes while
surface wave amplitudes from deep earthquakes and large (nuclear)
explosions are small at teleseismic distances.
• At seismic stations or network centers the following parameter readings
should be obligatory during routine analysis: onset time and, if possible,
polarity of the first arriving phase; maximum P amplitude A [nm] and period
T [s]; onset time of secondary phases; and for shallow sources additionally
the maximum surface-wave amplitude A [µm] and period T [s].
• P-wave amplitudes for the determination of the short-period body-wave
magnitude mb have to be measured on standard short-period (WWSSN-SP
simulated) records in the period-range 0.5 s < T < 2 s whereas the surface-
wave amplitudes for the determination of the surface-wave magnitude Ms
have to be measured on standard long-period filtered (SRO-LP or WWSSN-
LP simulated) records, typically in the period range 17 s < T < 23 s.
• For more guidance on magnitude determination, using also other phases and
records/filters, consult section 3.2.1 and related annexes.
• Networks and arrays should additionally measure and report slownesses and
azimuths for P waves.
• For improved determination of epicentre distance, the measurement and
reporting of travel-time differences such as S–P, SS-P etc. are very
important, and for improved hypocenter determination additionally the
proper identification and reporting of depth phases such as pP, sP, sS and of
core reflections (PcP, ScP etc.).
• Picking and reporting of onset-times, amplitudes and periods of other
significant phases, including those not identified, are encouraged by IASPEI
within the technical and personnel facilities available at observatories and
analysis centers as being a useful contribution to global research. These
extended possibilities for parameter reporting are now well supported by the
recently adopted IASPEI Seismic Format (ISF), which is much more flexible
and comprehensive than the traditional Telegraphic Format (see 10.2 as well
as IS 10.1 and 10.2).
• For reporting of seismic phases (including onsets not identified) one should
exclusively use the new IASPEI phase names. For the definition of seismic
phases and their ray paths see IS 2.1.
Acknowledgments
The authors are very grateful to A. Douglas and R. D. Adams for their
careful reviews which helped to significantly improve the original draft.
Thanks go also to staff members of the Geophysical Survey of the Russian
Academy of Science in Obninsk who shared in reviewing the various
sections of this Chapter (Ye. A. Babkova, L. S. Čepkunas, I. P.
Gabsatarova, M. B. Kolomiyez, S. G. Poygina and V. D. Theophylaktov).
Many of their valuable suggestions and references to Russian experience in
seismogram analysis were taken into account.
Recommended overview readings (see References under Miscellaneous in Volume 2)

Kennett (2002)
Kulhanek (1990 and 2002)
Payo (1986)
Richards (2002)
Scherbaum (2001 and 2002)
Simon (1981)
Willmore (1979)

Data Analysis and Seismogram Interpretation

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Data Analysis and Seismogram Interpretation

Uploaded by

Copyright:

Available Formats

11

Data Analysis and Seismogram

• recognize the occurrence of an earthquake in a record;

In modern digital observatory practice these procedures are implemented in computer

Fig. 11.2 Frequency range of seismological interest.

Accordingly, we first give a general introduction to routine seismogram interpretation of

11.2 Criteria and parameters for routine seismogram analysis

11.2.2 Key parameters: Onset time, amplitude, period and polarity

To calculate ground motion amplitudes from record amplitudes, the frequency-dependent

11.2.3 Advanced wavelet parameter reporting from digital records

11.2.4 Criteria to be used for phase identification

However, theoretically calculated onset-times based on a global average model should

11.2.4.2 Amplitudes, dominating periods and waveforms

Fig. 11.8 Three-component BB-velocity record at station MOX of a mine collapse in

As compared to shallow crustal earthquakes, deep earthquakes of comparable magnitude are

Fig. 11.13 Time-compressed long-period filtered three-component seismogram (SRO-LP

11.2.4.4 Example for documenting and reporting of seismogram parameter readings

• first line: date, event identifier, analyst;

11.2.5.1 Discrimination between shallow and deep earthquakes

determination of fault-plane solutions. This has reduced significantly the number of

Fig. 11.18 Broadband vertical-component seismograms of a deep (h = 119 km) earthquake

11.2.5.2 Discrimination between natural earthquakes and man-made seismic events

Based on these systematic differences in frequency content and polarization, some

Fig. 11.21 Short-period records at station MOX a) of an underground nuclear explosion at

In short-period records of higher time resolution the difference in frequency content,

11.2.6.1 What is the best way of analyzing three-component seismograms?

Take interest! Be curious! Ask questions!

1. Are you NEAR (D < 20°) or TELESEISMIC (D > 20°)?

2. Is your D < 100° or D > 100° ?

3. Are you SHALLOW or DEEP (> 70 km)?

4. Is the first strong horizontal arrival S or SKS ?

11.2.6.2 Hypocenter location

• general event classification (near/far; shallow/deep; D < 100°/> 100° etc.);

11.2.7 Magnitude determination

11.2.8 Hypocenter location by means of network and array recordings

11.3 Routine signal processing of digital seismograms

11.3.1 Signal detection

11.3.2 Signal filtering, restitution and simulation

H(s) = N * Π (s – zi) / Π (s – pk) (11.5)

Fig. 11.27 Left: Displacement amplitude response characteristics of classical seismographs;

Seismograph Zeros Poles

Seismograph Zeros Poles

H(s) = N * s3 / (s-p1)(s-p2)(s-p3)(s-p4) (11.6)

ωl2 = 25.27 with fl = 0.80 Hz and

Hfs (s) = Hr -1(s) * Hs(s). (11.7)

Simulation- Zeros Poles

Fig. 11.28 shows a comparison of the original three-component BB-velocity record of an

Hrest(s) = Hs -1(s) * Hs(s). (11.8)

Fig. 11.28 3-component recordings at station WET (Wettzell) of a local earthquake at an

The following features are shown in Figs. 11.32a-d:

Fig. 11.32a (for explanation see text on page 48)

Fig. 11.32b As Fig. 11.32a however with short-period WWSSN-SP simulation.

Fig. 11.32d As Fig. 11.32a but for long-period SRO-LP simulation.

11.3.3 Signal coherency at networks and arrays

Fig. 11.33a (see figure caption below)

11.3.4 f-k and vespagram analysis

Fig. 11.34 Top: Simulated vertical-component WWSSN-LP seismograms from an

11.3.6 Polarization analysis

Determination of particle motion is included in most of the analysis software. For

11.4 Software for routine analysis

Main features of the program are:

Furthermore, the following tasks are implemented:

11.4.5 Other programs and ORFEUS software links

ORFEUS (http://orfeus.knmi.nl/) presents a comprehensive list and links to available software